import numpy as np
import seaborn as sns
import pylab as pyl
import pandas as pd
import matplotlib.pyplot as plt
from scipy import stats
from sklearn import tree
from boruta import BorutaPy
from sklearn.svm import SVC
from scipy.stats import zscore
from sklearn.impute import KNNImputer
from mixed_naive_bayes import MixedNB
from imblearn.over_sampling import SMOTE
from sklearn.feature_selection import RFE
from sklearn.metrics import roc_auc_score
from sklearn.metrics import roc_curve, auc
from imblearn.pipeline import make_pipeline
from sklearn.ensemble import VotingClassifier
from sklearn.preprocessing import LabelEncoder
from sklearn.ensemble import BaggingClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.linear_model import LinearRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import cross_val_score
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from imblearn.under_sampling import RandomUnderSampler
from sklearn.model_selection import cross_val_score, KFold
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.ensemble import GradientBoostingClassifier as GBC
from sklearn.metrics import accuracy_score, classification_report, confusion_matrix
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score

data = pd.read_csv("churn_data.csv")
data

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5000 entries, 0 to 4999
Data columns (total 18 columns):
 #   Column                      Non-Null Count  Dtype  
---  ------                      --------------  -----  
 0   churn                       5000 non-null   object 
 1   accountlength               4951 non-null   float64
 2   internationalplan           4950 non-null   object 
 3   voicemailplan               4950 non-null   object 
 4   numbervmailmessages         4950 non-null   float64
 5   totaldayminutes             4950 non-null   float64
 6   totaldaycalls               4950 non-null   float64
 7   totaldaycharge              4950 non-null   float64
 8   totaleveminutes             4950 non-null   float64
 9   totalevecalls               4950 non-null   float64
 10  totalevecharge              4950 non-null   float64
 11  totalnightminutes           4950 non-null   float64
 12  totalnightcalls             4950 non-null   float64
 13  totalnightcharge            4950 non-null   float64
 14  totalintlminutes            4950 non-null   float64
 15  totalintlcalls              4950 non-null   float64
 16  totalintlcharge             4950 non-null   float64
 17  numbercustomerservicecalls  4950 non-null   float64
dtypes: float64(15), object(3)
memory usage: 703.3+ KB

print(str(data.duplicated().value_counts()))

False    5000
Name: count, dtype: int64

M = data.isna().sum()
M

churn                          0
accountlength                 49
internationalplan             50
voicemailplan                 50
numbervmailmessages           50
totaldayminutes               50
totaldaycalls                 50
totaldaycharge                50
totaleveminutes               50
totalevecalls                 50
totalevecharge                50
totalnightminutes             50
totalnightcalls               50
totalnightcharge              50
totalintlminutes              50
totalintlcalls                50
totalintlcharge               50
numbercustomerservicecalls    50
dtype: int64

M.sum()

849

data.count()/data.shape[0]

churn                         1.0000
accountlength                 0.9902
internationalplan             0.9900
voicemailplan                 0.9900
numbervmailmessages           0.9900
totaldayminutes               0.9900
totaldaycalls                 0.9900
totaldaycharge                0.9900
totaleveminutes               0.9900
totalevecalls                 0.9900
totalevecharge                0.9900
totalnightminutes             0.9900
totalnightcalls               0.9900
totalnightcharge              0.9900
totalintlminutes              0.9900
totalintlcalls                0.9900
totalintlcharge               0.9900
numbercustomerservicecalls    0.9900
dtype: float64

missing_values = data[data.isnull().any(axis=1)]
missing_values

data.describe()

fig, ax = plt.subplots(3, 5, figsize=(15, 10))
variables = ['accountlength', 'numbervmailmessages', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge']
indice = 0
for i in variables: 
    data[i].plot.box(ax=ax[0, indice])
    data[i].plot.density(ax=ax[1, indice], legend=True)
    data[i].plot.hist(ax=ax[2, indice], label=i, bins=10, legend=True)
    indice += 1
    
plt.tight_layout()

fig, ax = plt.subplots(3, 6, figsize=(15, 10))
variables = ['totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes' , 'totalnightcalls', 'totalnightcharge']
indice = 0
for i in variables: 
    data[i].plot.box(ax=ax[0, indice])
    data[i].plot.density(ax=ax[1, indice], legend=True)
    data[i].plot.hist(ax=ax[2, indice], label=i, bins=10, legend=True)
    indice += 1
    
plt.tight_layout()

fig, ax = plt.subplots(3, 4, figsize=(15, 10))
variables = ['totalintlminutes','totalintlcalls', 'totalintlcharge', 'numbercustomerservicecalls']
indice = 0
for i in variables: 
    data[i].plot.box(ax=ax[0, indice])
    data[i].plot.density(ax=ax[1, indice], legend=True)
    data[i].plot.hist(ax=ax[2, indice], label=i, bins=10, legend=True)
    indice += 1
    
plt.tight_layout()

variables = ['churn', 'internationalplan', 'voicemailplan']
fig, ax = plt.subplots(1, 3, figsize=(15, 5))

titles = ['Churn', 'International Plan', 'Voicemail Plan']

for i, variable in enumerate(variables):
    data[variable].value_counts().plot(kind='bar', ax=ax[i])
    ax[i].set_title(titles[i]) 

plt.tight_layout()

variables = ['accountlength', 'numbervmailmessages', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
                   'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
                   'totalnightcharge', 'totalintlminutes', 'totalintlcalls', 'totalintlcharge',
                   'numbercustomerservicecalls']

num_rows = 3 
num_cols = 5 
fig, axes = plt.subplots(nrows=num_rows, ncols=num_cols, figsize=(15, 15))

for i, variable in enumerate(variables):
    row = i // num_cols  
    col = i % num_cols  
    ax = data[[variable]].boxplot(ax=axes[row, col])
    q3 = data[variable].quantile(0.75)
    q1 = data[variable].quantile(0.25)
    iqr = q3 - q1
    topcut = q3 + 1.5 * iqr
    lowcut = q1 - 1.5 * iqr
    ax.axhline(topcut, c='red', ls='--')
    ax.axhline(lowcut, c='orange', ls='--')


plt.tight_layout()

variables = ['accountlength', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
             'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
             'totalnightcharge', 'totalintlminutes', 'totalintlcharge']

def detect_outliers_iqr(data):
    Q1 = data.quantile(0.25)
    Q3 = data.quantile(0.75)
    IQR = Q3 - Q1
    lower_bound = Q1 - 1.5 * IQR
    upper_bound = Q3 + 1.5 * IQR
    return data[(data < lower_bound) | (data > upper_bound)]

def detect_outliers_mad(data, threshold=3):
    median = data.median()
    abs_deviations = np.abs(data - median)
    mad = abs_deviations.median()
    return data[abs_deviations > threshold * mad]

outliers_iqr = {}
outliers_mad = {}
outliers_z = {}

for var in variables:
    outliers_iqr[var] = set(detect_outliers_iqr(data[var]).index)
    outliers_mad[var] = set(detect_outliers_mad(data[var]).index)
    
    mean = data[var].mean()
    std_dev = data[var].std()
    z_scores = (data[var] - mean) / std_dev
    outliers_z[var] = set(data[abs(z_scores) > 3].index)

common_outliers = {}
for var in variables:
    common_outliers[var] = outliers_iqr[var].intersection(outliers_mad[var], outliers_z[var])

for var in variables:
    print(f"Common outliers for '{var}':")
    print(common_outliers[var])
    print(f"Number of common outliers for '{var}': {len(common_outliers[var])}")
    print("\n")

Common outliers for 'accountlength':
{416, 1408, 4260, 4389, 4395, 1551, 3216, 817, 1886, 1751, 4379, 4798}
Number of common outliers for 'accountlength': 12


Common outliers for 'totaldayminutes':
{1345, 1986, 2594, 2753, 2252, 365, 1052, 2736, 1397, 4981, 985, 3993, 4892}
Number of common outliers for 'totaldayminutes': 13


Common outliers for 'totaldaycalls':
{1345, 1121, 740, 1989, 4049, 4562, 3187, 692, 468, 1397, 1460, 1144, 4155, 4831}
Number of common outliers for 'totaldaycalls': 14


Common outliers for 'totaldaycharge':
{1345, 1986, 2594, 2753, 2252, 365, 1052, 2736, 1397, 4981, 985, 3993, 4892}
Number of common outliers for 'totaldaycharge': 13


Common outliers for 'totaleveminutes':
{32, 2551, 1960, 3996, 3370, 2732, 1233, 2932, 821, 3989, 4375, 4885, 889, 2331, 4892}
Number of common outliers for 'totaleveminutes': 15


Common outliers for 'totalevecalls':
{960, 3940, 4741, 646, 301, 1615, 3219, 2932, 58}
Number of common outliers for 'totalevecalls': 9


Common outliers for 'totalevecharge':
{32, 2551, 3996, 3370, 2732, 1233, 2932, 821, 533, 4375, 3989, 889, 4885, 2331, 4892}
Number of common outliers for 'totalevecharge': 15


Common outliers for 'totalnightminutes':
{3107, 4707, 1317, 1445, 2663, 1260, 3247, 3376, 2321, 883, 244, 3060, 1238, 4727, 1113, 922, 3899, 3807}
Number of common outliers for 'totalnightminutes': 18


Common outliers for 'totalnightcalls':
{2659, 4707, 3560, 3211, 2988, 493, 3895, 3535, 2288, 4207, 3570, 2903, 3799, 4410, 4508}
Number of common outliers for 'totalnightcalls': 15


Common outliers for 'totalnightcharge':
{3107, 4707, 1317, 1445, 2663, 1260, 3247, 3376, 2321, 883, 244, 3060, 1238, 4727, 1113, 922, 3899, 3807}
Number of common outliers for 'totalnightcharge': 18


Common outliers for 'totalintlminutes':
{1028, 4371, 4504, 1564, 2345, 4396, 2733, 179, 3891, 1080, 2362, 314, 3906, 4935, 712, 3400, 3275, 3403, 3916, 3663, 2513, 595, 4949, 343, 2906, 3290, 4451, 2918, 488, 2669, 878, 3568, 4976, 115, 1400, 762}
Number of common outliers for 'totalintlminutes': 36


Common outliers for 'totalintlcharge':
{1028, 4371, 4504, 1564, 2345, 4396, 2733, 179, 3891, 1080, 2362, 314, 3906, 4935, 712, 3400, 3275, 3403, 3916, 3663, 2513, 595, 4949, 343, 2906, 3290, 4451, 2918, 488, 2669, 878, 3568, 4976, 115, 1400, 762}
Number of common outliers for 'totalintlcharge': 36

variables = ['numbervmailmessages', 'totalintlcalls', 'numbercustomerservicecalls']

def detect_outliers_iqr(data):
    Q1 = data.quantile(0.25)
    Q3 = data.quantile(0.75)
    IQR = Q3 - Q1
    lower_bound = Q1 - 1.5 * IQR
    upper_bound = Q3 + 1.5 * IQR
    return data[(data < lower_bound) | (data > upper_bound)]

def detect_outliers_mad(data, threshold=3):
    median = data.median()
    abs_deviations = np.abs(data - median)
    mad = abs_deviations.median()
    return data[abs_deviations > threshold * mad]

outliers_iqr = {}
outliers_mad = {}

for var in variables:
    outliers_iqr[var] = set(detect_outliers_iqr(data[var]).index)
    outliers_mad[var] = set(detect_outliers_mad(data[var]).index)
    
common_outliers = {}
for var in variables:
    common_outliers[var] = outliers_iqr[var].intersection(outliers_mad[var])

for var in variables:
    print(f"Common outliers for '{var}':")
    print(common_outliers[var])
    print(f"Number of common outliers for '{var}': {len(common_outliers[var])}")
    print("\n")

Common outliers for 'numbervmailmessages':
{1797, 1285, 2570, 268, 3342, 4369, 4625, 277, 149, 790, 4502, 2457, 2716, 1441, 4516, 423, 4777, 4778, 1454, 3246, 2608, 4275, 1846, 1596, 2366, 1602, 3523, 1732, 3524, 3910, 71, 2887, 4933, 845, 1487, 2768, 3279, 3154, 599, 3543, 4824, 4059, 3165, 4445, 1378, 4706, 3174, 615, 872, 3433, 4713, 4588, 4717, 3826, 4338, 1908, 1269, 3577, 4859, 2686}
Number of common outliers for 'numbervmailmessages': 60


Common outliers for 'totalintlcalls':
{514, 3587, 2562, 4108, 2576, 22, 3097, 4126, 1567, 4641, 3109, 4134, 41, 4652, 1581, 3636, 3128, 2621, 1092, 4679, 588, 4684, 3664, 4693, 4184, 2156, 4726, 1657, 636, 642, 3206, 2703, 1168, 3727, 153, 1179, 3230, 674, 2212, 2724, 182, 185, 698, 4286, 4292, 211, 723, 2775, 3800, 219, 3291, 3310, 3826, 756, 4857, 250, 3839, 3840, 1797, 2826, 4877, 272, 2835, 4386, 1832, 1333, 4917, 4408, 3386, 2883, 837, 329, 842, 1355, 341, 854, 3418, 347, 4955, 863, 1889, 3938, 1392, 4465, 2930, 883, 377, 3454, 2947, 3973, 4999, 1419, 2956, 911, 3991, 921, 2970, 420, 1960, 3516, 957, 3536, 2001, 3025, 3541, 982, 4568, 474, 2018, 483, 1526, 504, 1021, 3071}
Number of common outliers for 'totalintlcalls': 114


Common outliers for 'numbercustomerservicecalls':
{3585, 1538, 3079, 3081, 522, 3595, 1038, 4110, 4114, 4628, 21, 4631, 542, 2592, 4130, 547, 4132, 4644, 3112, 48, 54, 3131, 3643, 4670, 3140, 3655, 3144, 588, 4687, 3665, 4692, 2646, 1121, 1638, 3692, 1133, 3181, 1142, 3190, 635, 1150, 1662, 4740, 3717, 136, 1673, 655, 3228, 1694, 4772, 1193, 2218, 1707, 2732, 3243, 4778, 2223, 1713, 181, 694, 4790, 2248, 721, 4307, 1241, 4315, 1246, 736, 2785, 2786, 742, 235, 2283, 1273, 4346, 3323, 771, 1284, 778, 1802, 2827, 3851, 2322, 2327, 293, 1831, 1325, 4406, 4919, 1851, 4417, 841, 1865, 332, 2380, 3918, 2387, 3413, 2903, 3421, 2402, 874, 2415, 4982, 1912, 2428, 1919, 1407, 3973, 902, 392, 2952, 2953, 4489, 908, 2444, 2958, 911, 2960, 2961, 3468, 3983, 3487, 2979, 1970, 4018, 1973, 1974, 3509, 4023, 4537, 2493, 3009, 3017, 4044, 974, 3026, 2515, 4562, 3549, 1502, 2031, 4085, 2553, 509}
Number of common outliers for 'numbercustomerservicecalls': 145

data[['totaldayminutes', 'totaldaycalls',	'totaldaycharge']].corr()

data[['totaleveminutes', 'totalevecalls',	'totalevecharge']].corr()

data[['totalnightminutes', 'totalnightcalls',	'totalnightcharge']].corr()

data[['totalintlminutes','totalintlcalls', 'totalintlcharge']].corr()

fig, axs = plt.subplots(1, 4, figsize=(17, 5))

data.plot.scatter('totaldayminutes', 'totaldaycharge', ax=axs[0])
axs[0].set_title('Total day minutes vs Total day charge')

data.plot.scatter('totaleveminutes', 'totalevecharge', ax=axs[1])
axs[1].set_title('Total evening minutes vs Total evening charge')

data.plot.scatter('totalnightminutes', 'totalnightcharge', ax=axs[2])
axs[2].set_title('Total night minutes vs Total night charge')

data.plot.scatter('totalintlminutes', 'totalintlcharge', ax=axs[3])
axs[3].set_title('Total international minutes vs Total international charge')

plt.tight_layout()
plt.show()

columns_to_encode = ['churn', 'internationalplan', 'voicemailplan']

label_encoder = LabelEncoder()

for col in columns_to_encode:
    data[col] = label_encoder.fit_transform(data[col])

data

for row in data.itertuples():
    if data.isnull().sum(axis=1)[row[0]] >= 2:
        print(f"Row {row[0]} has {data.isnull().sum(axis=1)[row[0]]} NaN values")

Row 233 has 2 NaN values
Row 376 has 2 NaN values
Row 485 has 2 NaN values
Row 517 has 2 NaN values
Row 546 has 2 NaN values
Row 592 has 2 NaN values
Row 594 has 2 NaN values
Row 717 has 2 NaN values
Row 748 has 2 NaN values
Row 1082 has 2 NaN values
Row 1091 has 2 NaN values
Row 1178 has 2 NaN values
Row 1351 has 2 NaN values
Row 1437 has 2 NaN values
Row 1737 has 2 NaN values
Row 1839 has 2 NaN values
Row 1984 has 2 NaN values
Row 2095 has 2 NaN values
Row 2107 has 2 NaN values
Row 2246 has 2 NaN values
Row 2324 has 2 NaN values
Row 2524 has 2 NaN values
Row 2754 has 2 NaN values
Row 2762 has 2 NaN values
Row 2944 has 2 NaN values
Row 2947 has 2 NaN values
Row 3154 has 2 NaN values
Row 3262 has 2 NaN values
Row 3633 has 2 NaN values
Row 3686 has 3 NaN values
Row 3752 has 2 NaN values
Row 3791 has 3 NaN values
Row 3937 has 2 NaN values
Row 3967 has 2 NaN values
Row 4004 has 2 NaN values
Row 4071 has 2 NaN values
Row 4081 has 2 NaN values
Row 4147 has 2 NaN values
Row 4215 has 2 NaN values
Row 4396 has 2 NaN values
Row 4612 has 2 NaN values
Row 4727 has 2 NaN values
Row 4731 has 2 NaN values

variables_to_fill = ['totaldaycalls', 'totalevecalls', 'totalnightcalls','totalintlcalls']

data_mean_imputed = data.copy()
data_median_imputed = data.copy()

for col in variables_to_fill:
    data_mean_imputed[col].fillna(data[col].mean(), inplace=True)
    data_median_imputed[col].fillna(data[col].median(), inplace=True)

data_interpolated = data.copy()
data_interpolated[variables_to_fill] = data_interpolated[variables_to_fill].interpolate(method='linear')

imputer = KNNImputer(n_neighbors=5)
data_knn_imputed = data.copy()
data_knn_imputed[variables_to_fill] = imputer.fit_transform(data_knn_imputed[variables_to_fill])

mean_imputed_stats = data_mean_imputed[variables_to_fill].describe()
median_imputed_stats = data_median_imputed[variables_to_fill].describe()
interpolated_stats = data_interpolated[variables_to_fill].describe()
knn_imputed_stats = data_knn_imputed[variables_to_fill].describe()

without_imputation = data[['totaldaycalls', 'totalevecalls','totalnightcalls','totalintlcalls']].describe()
print("Statistics without Imputation:")
print(without_imputation)

print("\nStatistics for Mean Imputation:")
print(mean_imputed_stats)

print("\nStatistics for Median Imputation:")
print(median_imputed_stats)

print("\nStatistics for Linear Interpolation:")
print(interpolated_stats)

print("\nStatistics for KNN Imputation:")
print(knn_imputed_stats)

Statistics without Imputation:
       totaldaycalls  totalevecalls  totalnightcalls  totalintlcalls
count    4950.000000    4950.000000      4950.000000     4950.000000
mean      100.038788     100.243838        99.932929        4.432525
std        19.844529      19.837380        19.939450        2.448457
min         0.000000       0.000000         0.000000        0.000000
25%        87.000000      87.000000        87.000000        3.000000
50%       100.000000     101.000000       100.000000        4.000000
75%       113.000000     114.000000       113.000000        6.000000
max       165.000000     170.000000       175.000000       20.000000

Statistics for Mean Imputation:
       totaldaycalls  totalevecalls  totalnightcalls  totalintlcalls
count    5000.000000    5000.000000      5000.000000     5000.000000
mean      100.038788     100.243838        99.932929        4.432525
std        19.745037      19.737924        19.839483        2.436181
min         0.000000       0.000000         0.000000        0.000000
25%        87.000000      87.000000        87.000000        3.000000
50%       100.000000     100.243838       100.000000        4.000000
75%       113.000000     113.000000       113.000000        6.000000
max       165.000000     170.000000       175.000000       20.000000

Statistics for Median Imputation:
       totaldaycalls  totalevecalls  totalnightcalls  totalintlcalls
count    5000.000000    5000.000000      5000.000000     5000.000000
mean      100.038400     100.251400        99.933600        4.428200
std        19.745037      19.738068        19.839484        2.436562
min         0.000000       0.000000         0.000000        0.000000
25%        87.000000      87.000000        87.000000        3.000000
50%       100.000000     101.000000       100.000000        4.000000
75%       113.000000     113.000000       113.000000        6.000000
max       165.000000     170.000000       175.000000       20.000000

Statistics for Linear Interpolation:
       totaldaycalls  totalevecalls  totalnightcalls  totalintlcalls
count    5000.000000    5000.000000      5000.000000     5000.000000
mean      100.021300     100.240400        99.940700        4.433500
std        19.797851      19.774226        19.904891        2.443263
min         0.000000       0.000000         0.000000        0.000000
25%        87.000000      87.000000        87.000000        3.000000
50%       100.000000     100.500000       100.000000        4.000000
75%       113.000000     114.000000       113.000000        6.000000
max       165.000000     170.000000       175.000000       20.000000

Statistics for KNN Imputation:
       totaldaycalls  totalevecalls  totalnightcalls  totalintlcalls
count     5000.00000    5000.000000      5000.000000      5000.00000
mean       100.02932     100.246000        99.939800         4.43208
std         19.76802      19.759962        19.855006         2.43898
min          0.00000       0.000000         0.000000         0.00000
25%         87.00000      87.000000        87.000000         3.00000
50%        100.00000     100.000000       100.000000         4.00000
75%        113.00000     113.050000       113.000000         6.00000
max        165.00000     170.000000       175.000000        20.00000

for col in variables_to_fill:
    plt.figure(figsize=(14, 4))

    plt.subplot(1, 5, 1)
    sns.histplot(data[col], bins=20, kde=True, color='purple', label='Without Imputation')
    plt.legend()
    
    plt.subplot(1, 5, 2)
    sns.histplot(data_mean_imputed[col], bins=20, kde=True, color='blue', label='Mean Imputation')
    plt.legend()
    
    plt.subplot(1, 5, 3)
    sns.histplot(data_median_imputed[col], bins=20, kde=True, color='orange', label='Median Imputation')
    plt.legend()

    plt.subplot(1, 5, 4)
    sns.histplot(data_interpolated[col], bins=20, kde=True, color='green', label='Linear Interpolation')
    plt.legend()

    plt.subplot(1, 5, 5)
    sns.histplot(data_knn_imputed[col], bins=20, kde=True, color='red', label='KNN Imputation')
    plt.legend()

    plt.tight_layout()

imputer = KNNImputer(n_neighbors=5)
data_imputed = imputer.fit_transform(data)

data = pd.DataFrame(data_imputed, columns=data.columns)

imputed_stats = data.describe()
print("\nStatistics for KNN Imputation:")
print(imputed_stats)

Statistics for KNN Imputation:
             churn  accountlength  internationalplan  voicemailplan  \
count  5000.000000    5000.000000        5000.000000    5000.000000   
mean      0.141400     100.226784           0.113800       0.282400   
std       0.348469      39.554740           0.347669       0.471905   
min       0.000000       1.000000           0.000000       0.000000   
25%       0.000000      73.000000           0.000000       0.000000   
50%       0.000000     100.000000           0.000000       0.000000   
75%       0.000000     127.000000           0.000000       1.000000   
max       1.000000     243.000000           2.000000       2.000000   

       numbervmailmessages  totaldayminutes  totaldaycalls  totaldaycharge  \
count          5000.000000      5000.000000    5000.000000     5000.000000   
mean              7.760680       180.303714     100.035360       30.638822   
std              13.499377        53.725615      19.766022        9.148733   
min               0.000000         0.000000       0.000000        0.000000   
25%               0.000000       143.900000      87.000000       24.430000   
50%               0.000000       180.050000     100.000000       30.600000   
75%              16.000000       216.000000     113.000000       36.750000   
max              52.000000       351.500000     165.000000       59.760000   

       totaleveminutes  totalevecalls  totalevecharge  totalnightminutes  \
count      5000.000000    5000.000000     5000.000000        5000.000000   
mean        200.634184     100.250200       17.054007         200.468532   
std          50.291279      19.759087        4.293597          50.276816   
min           0.000000       0.000000        0.000000           0.000000   
25%         166.800000      87.000000       14.140000         167.300000   
50%         201.000000     101.000000       17.090000         200.500000   
75%         233.800000     113.000000       19.900000         234.125000   
max         363.700000     170.000000       30.910000         395.000000   

       totalnightcalls  totalnightcharge  totalintlminutes  totalintlcalls  \
count       5000.00000       5000.000000       5000.000000      5000.00000   
mean          99.92300          9.017016         10.262268         4.42888   
std           19.85791          2.271356          2.752844         2.43862   
min            0.00000          0.000000          0.000000         0.00000   
25%           87.00000          7.510000          8.500000         3.00000   
50%          100.00000          9.020000         10.300000         4.00000   
75%          113.00000         10.557000         12.000000         6.00000   
max          175.00000         17.770000         20.000000        20.00000   

       totalintlcharge  numbercustomerservicecalls  
count      5000.000000                  5000.00000  
mean          2.771706                     1.56956  
std           0.741653                     1.29969  
min           0.000000                     0.00000  
25%           2.309000                     1.00000  
50%           2.780000                     1.00000  
75%           3.240000                     2.00000  
max           5.400000                     9.00000

data_out_rem = data.copy()

variables = ['accountlength', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
             'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes',
             'totalnightcalls', 'totalnightcharge', 'totalintlminutes', 'totalintlcharge']

def detect_outliers_iqr(data):
    Q1 = data.quantile(0.25)
    Q3 = data.quantile(0.75)
    IQR = Q3 - Q1
    lower_bound = Q1 - 1.5 * IQR
    upper_bound = Q3 + 1.5 * IQR
    return data[(data < lower_bound) | (data > upper_bound)]

def detect_outliers_mad(data, threshold=3):
    median = data.median()
    abs_deviations = np.abs(data - median)
    mad = abs_deviations.median()
    return data[abs_deviations > threshold * mad]

outliers_iqr = {}
outliers_mad = {}
outliers_z = {}

for var in variables:
    outliers_iqr[var] = set(detect_outliers_iqr(data[var]).index)
    outliers_mad[var] = set(detect_outliers_mad(data[var]).index)
    
    mean = data[var].mean()
    std_dev = data[var].std()
    z_scores = (data[var] - mean) / std_dev
    outliers_z[var] = set(data[abs(z_scores) > 3].index)

common_outliers = {}
for var in variables:
    common_outliers[var] = outliers_iqr[var].intersection(outliers_mad[var], outliers_z[var])

common_indices = set.intersection(*common_outliers.values())
data_out_rem = data_out_rem.drop(common_indices)

for var in variables:
    common_var_outliers = common_outliers[var].intersection(data_out_rem.index)
    print(f"Number of common outliers removed for '{var}': {len(common_var_outliers)}")

stats_after = {}
for var in variables:
    stats_after[var] = data_out_rem[~data_out_rem.index.isin(common_outliers[var])][var].describe()

print("\nStatistics After Removing Outliers:")
for var, stats in stats_after.items():
    print(f"\nStatistics for '{var}':")
    print(stats)

Number of common outliers removed for 'accountlength': 12
Number of common outliers removed for 'totaldayminutes': 14
Number of common outliers removed for 'totaldaycalls': 14
Number of common outliers removed for 'totaldaycharge': 13
Number of common outliers removed for 'totaleveminutes': 17
Number of common outliers removed for 'totalevecalls': 9
Number of common outliers removed for 'totalevecharge': 15
Number of common outliers removed for 'totalnightminutes': 18
Number of common outliers removed for 'totalnightcalls': 15
Number of common outliers removed for 'totalnightcharge': 18
Number of common outliers removed for 'totalintlminutes': 39
Number of common outliers removed for 'totalintlcharge': 39

Statistics After Removing Outliers:

Statistics for 'accountlength':
count    4988.000000
mean       99.918388
std        39.097257
min         1.000000
25%        73.000000
50%       100.000000
75%       126.000000
max       217.000000
Name: accountlength, dtype: float64

Statistics for 'totaldayminutes':
count    4986.000000
mean      180.514054
std        53.030303
min        19.500000
25%       144.100000
50%       180.100000
75%       216.000000
max       338.400000
Name: totaldayminutes, dtype: float64

Statistics for 'totaldaycalls':
count    4986.000000
mean      100.127517
std        19.447740
min        42.000000
25%        87.000000
50%       100.000000
75%       113.000000
max       158.000000
Name: totaldaycalls, dtype: float64

Statistics for 'totaldaycharge':
count    4987.000000
mean       30.669035
std         9.038280
min         3.210000
25%        24.450000
50%        30.600000
75%        36.750000
max        57.530000
Name: totaldaycharge, dtype: float64

Statistics for 'totaleveminutes':
count    4983.000000
mean      200.806125
std        49.487535
min        52.900000
25%       167.000000
50%       201.000000
75%       233.800000
max       350.900000
Name: totaleveminutes, dtype: float64

Statistics for 'totalevecalls':
count    4991.000000
mean      100.271889
std        19.531729
min        42.000000
25%        87.000000
50%       101.000000
75%       113.000000
max       159.000000
Name: totalevecalls, dtype: float64

Statistics for 'totalevecharge':
count    4985.000000
mean       17.073894
std         4.232246
min         4.180000
25%        14.170000
50%        17.090000
75%        19.900000
max        29.930000
Name: totalevecharge, dtype: float64

Statistics for 'totalnightminutes':
count    4982.000000
mean      200.266792
std        49.353499
min        50.100000
25%       167.325000
50%       200.400000
75%       233.700000
max       350.200000
Name: totalnightminutes, dtype: float64

Statistics for 'totalnightcalls':
count    4985.000000
mean       99.917553
std        19.517249
min        41.000000
25%        87.000000
50%       100.000000
75%       113.000000
max       159.000000
Name: totalnightcalls, dtype: float64

Statistics for 'totalnightcharge':
count    4982.000000
mean        9.007922
std         2.229986
min         2.250000
25%         7.520000
50%         9.010000
75%        10.537500
max        15.760000
Name: totalnightcharge, dtype: float64

Statistics for 'totalintlminutes':
count    4961.000000
mean       10.309825
std         2.623528
min         2.100000
25%         8.600000
50%        10.400000
75%        12.000000
max        18.500000
Name: totalintlminutes, dtype: float64

Statistics for 'totalintlcharge':
count    4961.000000
mean        2.783654
std         0.706657
min         0.570000
25%         2.320000
50%         2.810000
75%         3.240000
max         4.970000
Name: totalintlcharge, dtype: float64

data_out_rem = data.copy()

variables = ['numbervmailmessages', 'totalintlcalls', 'numbercustomerservicecalls']

def detect_outliers_iqr(data):
    Q1 = data.quantile(0.25)
    Q3 = data.quantile(0.75)
    IQR = Q3 - Q1
    lower_bound = Q1 - 1.5 * IQR
    upper_bound = Q3 + 1.5 * IQR
    return data[(data < lower_bound) | (data > upper_bound)]

def detect_outliers_mad(data, threshold=4):
    median = data.median()
    abs_deviations = np.abs(data - median)
    mad = abs_deviations.median()
    return data[abs_deviations > threshold * mad]

outliers_iqr = {}
outliers_mad = {}

for var in variables:
    outliers_iqr[var] = set(detect_outliers_iqr(data[var]).index)
    outliers_mad[var] = set(detect_outliers_mad(data[var]).index)
    
common_outliers = {}
for var in variables:
    common_outliers[var] = outliers_iqr[var].intersection(outliers_mad[var])

common_indices = set.intersection(*common_outliers.values())
data_out_rem = data_out_rem.drop(common_indices)

for var in variables:
    common_var_outliers = common_outliers[var].intersection(data_out_rem.index)
    print(f"Number of common outliers removed for '{var}': {len(common_var_outliers)}")

stats_after = {}
for var in variables:
    stats_after[var] = data_out_rem[~data_out_rem.index.isin(common_outliers[var])][var].describe()

print("\nStatistics After Removing Outliers:")
for var, stats in stats_after.items():
    print(f"\nStatistics for '{var}':")
    print(stats)

Number of common outliers removed for 'numbervmailmessages': 99
Number of common outliers removed for 'totalintlcalls': 114
Number of common outliers removed for 'numbercustomerservicecalls': 51

Statistics After Removing Outliers:

Statistics for 'numbervmailmessages':
count    4901.000000
mean        7.031504
std        12.606129
min         0.000000
25%         0.000000
50%         0.000000
75%         8.800000
max        40.000000
Name: numbervmailmessages, dtype: float64

Statistics for 'totalintlcalls':
count    4886.000000
mean        4.232788
std         2.070127
min         0.000000
25%         3.000000
50%         4.000000
75%         5.000000
max        10.000000
Name: totalintlcalls, dtype: float64

Statistics for 'numbercustomerservicecalls':
count    4949.000000
mean        1.519256
std         1.205275
min         0.000000
25%         1.000000
50%         1.000000
75%         2.000000
max         5.000000
Name: numbercustomerservicecalls, dtype: float64

data.describe()

correlation_matrix = data.corr(numeric_only = True)
correlation_matrix

plt.figure(figsize=(12, 8))
sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm', fmt=".2f")
plt.show()

np.int = np.int32
np.float = np.float64
np.bool = np.bool_

X = data.drop('churn', axis=1)
y = data['churn']

forest = RandomForestClassifier(n_jobs=-1, class_weight='balanced', max_depth=5, random_state=42)

boruta_selector = BorutaPy(forest, n_estimators='auto', verbose=2, random_state=42)
boruta_selector.fit(X.values, y.values)

Iteration: 	1 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	2 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	3 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	4 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	5 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	6 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	7 / 100
Confirmed: 	0
Tentative: 	17
Rejected: 	0
Iteration: 	8 / 100
Confirmed: 	13
Tentative: 	0
Rejected: 	4


BorutaPy finished running.

Iteration: 	9 / 100
Confirmed: 	13
Tentative: 	0
Rejected: 	4

BorutaPy(estimator=RandomForestClassifier(class_weight='balanced', max_depth=5,
                                          n_estimators=116, n_jobs=-1,
                                          random_state=RandomState(MT19937) at 0x2A81CE80840),
         n_estimators='auto',
         random_state=RandomState(MT19937) at 0x2A81CE80840, verbose=2)

BorutaPy(estimator=RandomForestClassifier(class_weight='balanced', max_depth=5,
                                          n_estimators=116, n_jobs=-1,
                                          random_state=RandomState(MT19937) at 0x2A81CE80840),
         n_estimators='auto',
         random_state=RandomState(MT19937) at 0x2A81CE80840, verbose=2)

RandomForestClassifier(class_weight='balanced', max_depth=5, n_estimators=116,
                       n_jobs=-1,
                       random_state=RandomState(MT19937) at 0x2A81CE80840)

RandomForestClassifier(class_weight='balanced', max_depth=5, n_estimators=116,
                       n_jobs=-1,
                       random_state=RandomState(MT19937) at 0x2A81CE80840)

selected_features= X.columns[boruta_selector.support_].tolist()
removed_features= X.columns[~boruta_selector.support_].tolist()

print("Selected Features:", selected_features)
print("Removed Features:", removed_features)

Selected Features: ['internationalplan', 'voicemailplan', 'numbervmailmessages', 'totaldayminutes', 'totaldaycharge', 'totaleveminutes', 'totalevecharge', 'totalnightminutes', 'totalnightcharge', 'totalintlminutes', 'totalintlcalls', 'totalintlcharge', 'numbercustomerservicecalls']
Removed Features: ['accountlength', 'totaldaycalls', 'totalevecalls', 'totalnightcalls']

forest.fit(X.values, y.values)

feature_importances = forest.feature_importances_

feature_names = X.columns

sorted_idx = np.argsort(feature_importances)[::-1]

plt.figure(figsize=(12, 6))
plt.bar(range(X.shape[1]), feature_importances[sorted_idx], align="center")
plt.xticks(range(X.shape[1]), np.array(feature_names)[sorted_idx], rotation=90)
plt.xlabel("Feature")
plt.ylabel("Importance")
plt.title("Feature Importances after Boruta")
plt.show()

feature_range = range(9, len(X.columns) + 1,2)

cv_scores = []

for n in feature_range:
    rfe_selector = RFE(estimator=forest, n_features_to_select=n, step=1)
    scores = cross_val_score(rfe_selector, X, y, cv=5, scoring='accuracy')
    cv_scores.append(scores.mean())

plt.plot(feature_range, cv_scores)
plt.xlabel('Number of Features Selected')
plt.ylabel('Cross-Validation Accuracy')
plt.title('RFE Performance')
plt.show()

n_features_to_select = min(11, X.shape[1])

rfe_selector = RFE(estimator=forest, n_features_to_select=n_features_to_select, step=1)
rfe_selector.fit(X, y)

final_selected_features = X.columns[rfe_selector.support_]
removed_features = X.columns[~rfe_selector.support_]

feature_importances = rfe_selector.estimator_.feature_importances_
sorted_idx = np.argsort(feature_importances)[::-1]

plt.figure(figsize=(12, 6))
plt.bar(range(len(final_selected_features)), feature_importances[sorted_idx], align="center")
plt.xticks(range(len(final_selected_features)), np.array(final_selected_features)[sorted_idx], rotation=90)
plt.xlabel("Feature")
plt.ylabel("Importance")
plt.title("Feature Importances after RFE")
plt.show()

print("Final Selected Features:", final_selected_features)
print("Removed Features:", removed_features)

Final Selected Features: Index(['internationalplan', 'voicemailplan', 'numbervmailmessages',
       'totaldayminutes', 'totaldaycharge', 'totaleveminutes',
       'totalevecharge', 'totalintlminutes', 'totalintlcalls',
       'totalintlcharge', 'numbercustomerservicecalls'],
      dtype='object')
Removed Features: Index(['accountlength', 'totaldaycalls', 'totalevecalls', 'totalnightminutes',
       'totalnightcalls', 'totalnightcharge'],
      dtype='object')

accountlength_array = data['accountlength'].to_numpy()
pd.cut(accountlength_array, 10)

[(122.0, 146.2], (97.8, 122.0], (122.0, 146.2], (73.6, 97.8], (73.6, 97.8], ..., (49.4, 73.6], (146.2, 170.4], (49.4, 73.6], (97.8, 122.0], (73.6, 97.8]]
Length: 5000
Categories (10, interval[float64, right]): [(0.758, 25.2] < (25.2, 49.4] < (49.4, 73.6] < (73.6, 97.8] ... (146.2, 170.4] < (170.4, 194.6] < (194.6, 218.8] < (218.8, 243.0]]

totaldayminutes_array = data['totaldayminutes'].to_numpy()
pd.cut(totaldayminutes_array, 10)

[(246.05, 281.2], (140.6, 175.75], (210.9, 246.05], (281.2, 316.35], (140.6, 175.75], ..., (210.9, 246.05], (175.75, 210.9], (105.45, 140.6], (175.75, 210.9], (105.45, 140.6]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.352, 35.15] < (35.15, 70.3] < (70.3, 105.45] < (105.45, 140.6] ... (210.9, 246.05] < (246.05, 281.2] < (281.2, 316.35] < (316.35, 351.5]]

totaldaycalls_array = data['totaldaycalls'].to_numpy()
pd.cut(totaldaycalls_array, 10)

[(99.0, 115.5], (115.5, 132.0], (99.0, 115.5], (66.0, 82.5], (99.0, 115.5], ..., (115.5, 132.0], (82.5, 99.0], (82.5, 99.0], (66.0, 82.5], (99.0, 115.5]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.165, 16.5] < (16.5, 33.0] < (33.0, 49.5] < (49.5, 66.0] ... (99.0, 115.5] < (115.5, 132.0] < (132.0, 148.5] < (148.5, 165.0]]

totaldaycharge_array = data['totaldaycharge'].to_numpy()
pd.cut(totaldaycharge_array, 10)

[(41.832, 47.808], (23.904, 29.88], (35.856, 41.832], (47.808, 53.784], (23.904, 29.88], ..., (35.856, 41.832], (29.88, 35.856], (17.928, 23.904], (29.88, 35.856], (17.928, 23.904]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.0598, 5.976] < (5.976, 11.952] < (11.952, 17.928] < (17.928, 23.904] ... (35.856, 41.832] < (41.832, 47.808] < (47.808, 53.784] < (53.784, 59.76]]

totaleveminutes_array = data['totaleveminutes'].to_numpy()
pd.cut(totaleveminutes_array, 10)

[(181.85, 218.22], (181.85, 218.22], (109.11, 145.48], (36.37, 72.74], (145.48, 181.85], ..., (218.22, 254.59], (254.59, 290.96], (145.48, 181.85], (145.48, 181.85], (254.59, 290.96]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.364, 36.37] < (36.37, 72.74] < (72.74, 109.11] < (109.11, 145.48] ... (218.22, 254.59] < (254.59, 290.96] < (290.96, 327.33] < (327.33, 363.7]]

totalevecalls_array = data['totalevecalls'].to_numpy()
pd.cut(totalevecalls_array, 10)

[(85.0, 102.0], (102.0, 119.0], (102.0, 119.0], (85.0, 102.0], (119.0, 136.0], ..., (119.0, 136.0], (68.0, 85.0], (119.0, 136.0], (85.0, 102.0], (102.0, 119.0]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.17, 17.0] < (17.0, 34.0] < (34.0, 51.0] < (51.0, 68.0] ... (102.0, 119.0] < (119.0, 136.0] < (136.0, 153.0] < (153.0, 170.0]]

totalevecharge_array = data['totalevecharge'].to_numpy()
pd.cut(totalevecharge_array, 10)

[(15.455, 18.546], (15.455, 18.546], (9.273, 12.364], (3.091, 6.182], (12.364, 15.455], ..., (18.546, 21.637], (21.637, 24.728], (12.364, 15.455], (12.364, 15.455], (21.637, 24.728]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.0309, 3.091] < (3.091, 6.182] < (6.182, 9.273] < (9.273, 12.364] ... (18.546, 21.637] < (21.637, 24.728] < (24.728, 27.819] < (27.819, 30.91]]

totalnightminutes_array = data['totalnightminutes'].to_numpy()
pd.cut(totalnightminutes_array, 10)

[(237.0, 276.5], (237.0, 276.5], (158.0, 197.5], (158.0, 197.5], (158.0, 197.5], ..., (276.5, 316.0], (197.5, 237.0], (197.5, 237.0], (197.5, 237.0], (118.5, 158.0]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.395, 39.5] < (39.5, 79.0] < (79.0, 118.5] < (118.5, 158.0] ... (237.0, 276.5] < (276.5, 316.0] < (316.0, 355.5] < (355.5, 395.0]]

totalnightcalls_array = data['totalnightcalls'].to_numpy()
pd.cut(totalnightcalls_array, 10)

[(87.5, 105.0], (87.5, 105.0], (87.5, 105.0], (87.5, 105.0], (105.0, 122.5], ..., (105.0, 122.5], (105.0, 122.5], (87.5, 105.0], (87.5, 105.0], (87.5, 105.0]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.175, 17.5] < (17.5, 35.0] < (35.0, 52.5] < (52.5, 70.0] ... (105.0, 122.5] < (122.5, 140.0] < (140.0, 157.5] < (157.5, 175.0]]

totalnightcharge_array = data['totalnightcharge'].to_numpy()
pd.cut(totalnightcharge_array, 10)

[(10.662, 12.439], (10.662, 12.439], (7.108, 8.885], (7.108, 8.885], (7.108, 8.885], ..., (12.439, 14.216], (8.885, 10.662], (8.885, 10.662], (8.885, 10.662], (5.331, 7.108]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.0178, 1.777] < (1.777, 3.554] < (3.554, 5.331] < (5.331, 7.108] ... (10.662, 12.439] < (12.439, 14.216] < (14.216, 15.993] < (15.993, 17.77]]

totalintlminutes_array = data['totalintlminutes'].to_numpy()
pd.cut(totalintlminutes_array, 10)

[(8.0, 10.0], (12.0, 14.0], (12.0, 14.0], (6.0, 8.0], (10.0, 12.0], ..., (8.0, 10.0], (14.0, 16.0], (12.0, 14.0], (8.0, 10.0], (8.0, 10.0]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.02, 2.0] < (2.0, 4.0] < (4.0, 6.0] < (6.0, 8.0] ... (12.0, 14.0] < (14.0, 16.0] < (16.0, 18.0] < (18.0, 20.0]]

totalintlcharge_array = data['totalintlcharge'].to_numpy()
pd.cut(totalintlcharge_array, 10)

[(2.16, 2.7], (3.24, 3.78], (3.24, 3.78], (1.62, 2.16], (2.7, 3.24], ..., (2.16, 2.7], (3.78, 4.32], (3.24, 3.78], (2.16, 2.7], (2.16, 2.7]]
Length: 5000
Categories (10, interval[float64, right]): [(-0.0054, 0.54] < (0.54, 1.08] < (1.08, 1.62] < (1.62, 2.16] ... (3.24, 3.78] < (3.78, 4.32] < (4.32, 4.86] < (4.86, 5.4]]

pd.qcut(data['numbervmailmessages'], 10, duplicates='drop')

0         (24.0, 32.0]
1         (24.0, 32.0]
2       (-0.001, 24.0]
3       (-0.001, 24.0]
4       (-0.001, 24.0]
             ...      
4995      (32.0, 52.0]
4996    (-0.001, 24.0]
4997    (-0.001, 24.0]
4998    (-0.001, 24.0]
4999      (32.0, 52.0]
Name: numbervmailmessages, Length: 5000, dtype: category
Categories (3, interval[float64, right]): [(-0.001, 24.0] < (24.0, 32.0] < (32.0, 52.0]]

pd.qcut(data['numbervmailmessages'], 10, duplicates='drop').value_counts()

numbervmailmessages
(-0.001, 24.0]    4056
(24.0, 32.0]       499
(32.0, 52.0]       445
Name: count, dtype: int64

pd.qcut(data['totalintlcalls'], 10, duplicates='drop')

0          (2.0, 3.0]
1          (2.0, 3.0]
2          (4.0, 5.0]
3          (6.0, 8.0]
4          (2.0, 3.0]
            ...      
4995       (4.0, 5.0]
4996    (-0.001, 2.0]
4997       (3.0, 4.0]
4998       (5.0, 6.0]
4999      (8.0, 20.0]
Name: totalintlcalls, Length: 5000, dtype: category
Categories (7, interval[float64, right]): [(-0.001, 2.0] < (2.0, 3.0] < (3.0, 4.0] < (4.0, 5.0] < (5.0, 6.0] < (6.0, 8.0] < (8.0, 20.0]]

pd.qcut(data['totalintlcalls'], 10, duplicates='drop').value_counts()

totalintlcalls
(-0.001, 2.0]    1018
(2.0, 3.0]        993
(3.0, 4.0]        964
(4.0, 5.0]        715
(5.0, 6.0]        496
(6.0, 8.0]        480
(8.0, 20.0]       334
Name: count, dtype: int64

pd.qcut(data['numbercustomerservicecalls'], 10, duplicates='drop')

0       (-0.001, 1.0]
1       (-0.001, 1.0]
2       (-0.001, 1.0]
3          (1.0, 2.0]
4          (2.0, 3.0]
            ...      
4995       (1.0, 2.0]
4996       (2.0, 3.0]
4997    (-0.001, 1.0]
4998    (-0.001, 1.0]
4999    (-0.001, 1.0]
Name: numbercustomerservicecalls, Length: 5000, dtype: category
Categories (4, interval[float64, right]): [(-0.001, 1.0] < (1.0, 2.0] < (2.0, 3.0] < (3.0, 9.0]]

pd.qcut(data['numbercustomerservicecalls'], 10, duplicates='drop').value_counts()

numbercustomerservicecalls
(-0.001, 1.0]    2788
(1.0, 2.0]       1152
(2.0, 3.0]        668
(3.0, 9.0]        392
Name: count, dtype: int64

data_shuffled = data.sample(frac = 1, random_state=10)

data_shuffled

train_data, test_data = train_test_split(data_shuffled, test_size=0.2, random_state=10)

train_data

test_data

X_train = train_data.drop('churn', axis=1)
y_train = np.array(train_data['churn'])

X_test = test_data.drop('churn', axis=1)
y_test = np.array(test_data['churn'])

pipeline = make_pipeline(SMOTE(sampling_strategy=0.5, random_state=10), RandomUnderSampler(sampling_strategy=1.0, random_state=10))
X_train_balanced, y_train_balanced = pipeline.fit_resample(X_train,y_train)

X_train_balanced

scaler = StandardScaler()
X_train_balanced_scaled = scaler.fit_transform(X_train_balanced)
X_test_scaled = scaler.transform(X_test)

knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(X_train_balanced_scaled, y_train_balanced)

KNeighborsClassifier(n_neighbors=3)

KNeighborsClassifier(n_neighbors=3)

y_pred = knn.predict(X_test_scaled)

accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)

Accuracy: 0.81

k_values = [i for i in range (1,30)]
scores = []

scaler = StandardScaler()
X_train_balanced_scaled = scaler.fit_transform(X_train_balanced_scaled)

for k in k_values:
    knn = KNeighborsClassifier(n_neighbors=k)
    knn.fit(X_train_balanced_scaled, y_train_balanced)
    y_pred = knn.predict(X_test_scaled)
    score = accuracy_score(y_test, y_pred)
    scores.append(np.mean(score))

sns.lineplot(x = k_values, y = scores, marker = 'o')
plt.xlabel("K Values")
plt.ylabel("Accuracy Score")

Text(0, 0.5, 'Accuracy Score')

knn = KNeighborsClassifier(n_neighbors=9)
knn.fit(X_train_balanced_scaled, y_train_balanced)

y_pred = knn.predict(X_test_scaled)

accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)

Accuracy: 0.801

knn_b = KNeighborsClassifier(n_neighbors=9)
knn.fit(X_train_balanced_scaled, y_train_balanced)

y_pred = knn.predict(X_test_scaled)

accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)
    
print("Accuracy:", accuracy)
print("Precision:", precision)
print("Recall:", recall)
print("F1 Score:", f1)

Accuracy: 0.801
Precision: 0.40217391304347827
Recall: 0.7655172413793103
F1 Score: 0.5273159144893111

print("Classification Report for X_train_balanced data:")
print(classification_report(y_test, y_pred, zero_division=0))

correct_predictions = (y_pred == y_test)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

Classification Report for X_train_balanced data:
              precision    recall  f1-score   support

         0.0       0.95      0.81      0.87       855
         1.0       0.40      0.77      0.53       145

    accuracy                           0.80      1000
   macro avg       0.68      0.79      0.70      1000
weighted avg       0.87      0.80      0.82      1000

Number of correct predictions: 801

KNN_conf_matrix = confusion_matrix(y_test, y_pred)

plt.figure(figsize=(4, 4))
sns.heatmap(KNN_conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of KNN')
plt.show()

X_train_balanced = X_train_balanced.loc[:, X_train_balanced.columns != 'churn']
y_train_balanced = y_train_balanced.astype(int)

scaler_standard = StandardScaler()
numeric_features = ['accountlength', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
                    'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
                    'totalnightcharge', 'totalintlminutes', 'totalintlcharge']  

scaler_minmax = MinMaxScaler()
numeric_features_minmax = ['numbervmailmessages', 'totalintlcalls', 'numbercustomerservicecalls']  

X_numeric_train_balanced = X_train_balanced[numeric_features] 
X_numeric_train_balanced_scaled = scaler_standard.fit_transform(X_numeric_train_balanced)  
X_train_balanced_scaled = X_train_balanced.copy()
X_train_balanced_scaled[numeric_features] = X_numeric_train_balanced_scaled  

X_numeric_minmax = X_train_balanced_scaled[numeric_features_minmax]  
X_numeric_minmax_scaled = scaler_minmax.fit_transform(X_numeric_minmax)  
X_train_balanced_scaled[numeric_features_minmax] = X_numeric_minmax_scaled  

model_NB_b = MixedNB()

model_NB_b.fit(X_train_balanced_scaled, y_train_balanced)  

preds = model_NB_b.predict(X_train_balanced_scaled)

accuracy = accuracy_score(y_train_balanced, preds)
print(f"Accuracy on X_train_balanced data: {accuracy:.2f}")

print("Classification Report for X_train_balanced data:")
print(classification_report(y_train_balanced, preds, zero_division=0))

correct_predictions = (preds == y_train_balanced)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

conf_matrix = confusion_matrix(y_train_balanced, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of NB')
plt.show()

Accuracy on X_train_balanced data: 0.78
Classification Report for X_train_balanced data:
              precision    recall  f1-score   support

           0       0.81      0.74      0.78      1719
           1       0.76      0.82      0.79      1719

    accuracy                           0.78      3438
   macro avg       0.79      0.78      0.78      3438
weighted avg       0.79      0.78      0.78      3438

Number of correct predictions: 2696

model = DecisionTreeClassifier(random_state=0)
model.fit(X_train_balanced,y_train_balanced)

DecisionTreeClassifier(random_state=0)

DecisionTreeClassifier(random_state=0)

plt.subplots(nrows = 1,ncols = 1,figsize = (12,12), dpi=200)
tree.plot_tree(model, fontsize=5, filled=True)
plt.title('Decision Tree Not Prunned')
plt.show()

preds = model.predict(X_train_balanced)
(vals,ans)=np.unique(preds==y_train_balanced, return_counts=True)
ans[vals]/sum(ans)

array([1.])

def scores(model):
    preds = model.predict(X_test)
    print("Classification Report for test data:")
    print(classification_report(y_test, preds, zero_division=0))

scores(model)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.96      0.90      0.93       855
         1.0       0.58      0.79      0.67       145

    accuracy                           0.89      1000
   macro avg       0.77      0.84      0.80      1000
weighted avg       0.91      0.89      0.89      1000

max_depth=[]
accuracies = []
accuracies_train = []

for n in range(1,20,2):
    model = DecisionTreeClassifier(random_state=0,max_depth=n)
    model.fit(X_train_balanced,y_train_balanced)
    y_pred = model.predict(X_test)
    accuracy = accuracy_score(y_test, y_pred)
    y_pred_train = model.predict(X_train_balanced)
    accuracy_train =  accuracy_score(y_train_balanced, y_pred_train)
    accuracies_train.append(accuracy_train)
    max_depth.append(n)
    accuracies.append(accuracy)
    print(f"With Max Depth of {n} the accuracy on the train data is {accuracy_train} and the accuracy on the test data is {accuracy}")

With Max Depth of 1 the accuracy on the train data is 0.673065735892961 and the accuracy on the test data is 0.832
With Max Depth of 3 the accuracy on the train data is 0.8528214077952297 and the accuracy on the test data is 0.825
With Max Depth of 5 the accuracy on the train data is 0.9162303664921466 and the accuracy on the test data is 0.904
With Max Depth of 7 the accuracy on the train data is 0.9432809773123909 and the accuracy on the test data is 0.919
With Max Depth of 9 the accuracy on the train data is 0.9616055846422339 and the accuracy on the test data is 0.914
With Max Depth of 11 the accuracy on the train data is 0.9720767888307156 and the accuracy on the test data is 0.907
With Max Depth of 13 the accuracy on the train data is 0.9872018615474113 and the accuracy on the test data is 0.901
With Max Depth of 15 the accuracy on the train data is 0.997673065735893 and the accuracy on the test data is 0.894
With Max Depth of 17 the accuracy on the train data is 1.0 and the accuracy on the test data is 0.886
With Max Depth of 19 the accuracy on the train data is 1.0 and the accuracy on the test data is 0.886

plt.figure(figsize=(10, 6))
plt.plot(max_depth, accuracies_train, label='Train Accuracy', marker='o')
plt.plot(max_depth, accuracies, label='Test Accuracy', marker='o')
plt.title('Decision Tree Accuracy vs. Max Depth')
plt.xlabel('Max Depth')
plt.ylabel('Accuracy')
plt.legend()
plt.grid(True)
plt.show()

model = DecisionTreeClassifier(random_state=0,max_depth=7)
model.fit(X_train_balanced,y_train_balanced)

plt.subplots(nrows = 1,ncols = 1,figsize = (14,14), dpi=200)
tree.plot_tree(model, fontsize=5, filled=True)
plt.title('Decision Tree with Max Depth Value of 7')
plt.show()

model = DecisionTreeClassifier(random_state = 0)

param_grid = {
    'max_depth': [None,8,14,16],
    'min_samples_split': [2,5,10],
    'ccp_alpha':[0.0001,0.001,0.01]
}

grid_search = GridSearchCV(model, param_grid, cv=5)
grid_search.fit(X_train_balanced, y_train_balanced)

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 8, 14, 16],
                         'min_samples_split': [2, 5, 10]})

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 8, 14, 16],
                         'min_samples_split': [2, 5, 10]})

DecisionTreeClassifier(random_state=0)

DecisionTreeClassifier(random_state=0)

best_dt_model = grid_search.best_estimator_

best_dt_model

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, random_state=0)

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, random_state=0)

plt.subplots(nrows = 1,ncols = 1,figsize = (14,14), dpi=200)
tree.plot_tree(best_dt_model, fontsize=5, filled=True)
plt.title('Decision Tree with the best parameters defined from a Cross Validation Method')
plt.show()

scores(best_dt_model)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.97      0.93      0.95       855
         1.0       0.65      0.82      0.73       145

    accuracy                           0.91      1000
   macro avg       0.81      0.87      0.84      1000
weighted avg       0.92      0.91      0.91      1000

preds = best_dt_model.predict(X_test)
cm = confusion_matrix(y_test, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', cbar=True)
plt.xlabel('Predicted Labels')
plt.ylabel('True Labels')
plt.title('Confusion Matrix of Decision Tree Model')
plt.show()

model_f1 = DecisionTreeClassifier(random_state = 0)

grid_search_f1 = GridSearchCV(model_f1, param_grid, scoring='f1', cv=5)
grid_search_f1.fit(X_train_balanced, y_train_balanced)

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 8, 14, 16],
                         'min_samples_split': [2, 5, 10]},
             scoring='f1')

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 8, 14, 16],
                         'min_samples_split': [2, 5, 10]},
             scoring='f1')

DecisionTreeClassifier(random_state=0)

DecisionTreeClassifier(random_state=0)

best_dt_model_f1_b = grid_search_f1.best_estimator_

best_dt_model_f1_b

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, random_state=0)

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, random_state=0)

scores(best_dt_model_f1_b)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.97      0.93      0.95       855
         1.0       0.65      0.82      0.73       145

    accuracy                           0.91      1000
   macro avg       0.81      0.87      0.84      1000
weighted avg       0.92      0.91      0.91      1000

preds = best_dt_model_f1_b.predict(X_test)
cm = confusion_matrix(y_test, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', cbar=True)
plt.xlabel('Predicted Labels')
plt.ylabel('True Labels')
plt.title('Confusion Matrix of the F1 Tunned Decision Tree Model')
plt.show()

X_train_tune, X_train_final, y_train_tune, y_train_final = train_test_split(X_train_balanced, y_train_balanced, test_size=0.8, random_state=0)

bag = BaggingClassifier(random_state=0)

param_grid = {
    'n_estimators': [100, 150, 200, 250]
}

grid_search = GridSearchCV(bag, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_bag_b = grid_search.best_estimator_

best_bag_b

BaggingClassifier(n_estimators=200, random_state=0)

BaggingClassifier(n_estimators=200, random_state=0)

base_tree = DecisionTreeClassifier()

best_bag_b.fit(X_train_balanced, y_train_balanced)

BaggingClassifier(n_estimators=200, random_state=0)

BaggingClassifier(n_estimators=200, random_state=0)

rf = RandomForestClassifier(random_state=0)
param_grid = {
    'n_estimators': [100, 150, 200, 250],
    'max_features': [None, 'sqrt', 'log2']  
}

grid_search = GridSearchCV(rf, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_rf_b = grid_search.best_estimator_

best_rf_b

RandomForestClassifier(n_estimators=200, random_state=0)

RandomForestClassifier(n_estimators=200, random_state=0)

best_rf_b.fit(X_train_balanced, y_train_balanced)
scores(best_rf_b)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.97      0.96      0.96       855
         1.0       0.76      0.83      0.79       145

    accuracy                           0.94      1000
   macro avg       0.87      0.89      0.88      1000
weighted avg       0.94      0.94      0.94      1000

model =  AdaBoostClassifier(random_state=0)
param_grid = {
    'n_estimators': [100, 150, 200, 250]
}

grid_search = GridSearchCV(model, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_ada_b = grid_search.best_estimator_
best_ada_b

AdaBoostClassifier(n_estimators=100, random_state=0)

AdaBoostClassifier(n_estimators=100, random_state=0)

best_ada_b.fit(X_train_balanced, y_train_balanced)
scores(best_ada_b)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.94      0.89      0.91       855
         1.0       0.49      0.65      0.56       145

    accuracy                           0.85      1000
   macro avg       0.71      0.77      0.73      1000
weighted avg       0.87      0.85      0.86      1000

model =  GBC(random_state=0)
param_grid = {
    'n_estimators': [100, 150, 200, 250]
}

grid_search = GridSearchCV(model, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_gbc_b = grid_search.best_estimator_

best_gbc_b.fit(X_train_balanced, y_train_balanced)
scores(best_gbc_b)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.97      0.96      0.96       855
         1.0       0.77      0.82      0.80       145

    accuracy                           0.94      1000
   macro avg       0.87      0.89      0.88      1000
weighted avg       0.94      0.94      0.94      1000

models = [best_bag_b,best_rf_b, best_ada_b, best_gbc_b]
model_names = ['Bagging','Random Forest','Ada Booster','Gradient Boosting']

plt.figure(figsize=(8, 8))

for model, name in zip(models, model_names):
    y_pred_proba = model.predict_proba(X_test)[:, 1]
    fpr, tpr, _ = roc_curve(y_test, y_pred_proba)
    roc_auc = auc(fpr, tpr)
    plt.plot(fpr, tpr, lw=2, label=f'{name} (AUC = {roc_auc:.2f})')

plt.plot([0, 1], [0, 1], linestyle='--', color='gray', label='Random Classifier')

plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver Operating Characteristic (ROC) Curve')
plt.legend(loc='lower right')
plt.show()

X_train_balanced = X_train_balanced.loc[:, X_train_balanced.columns != 'churn']
y_train_balanced = y_train_balanced.astype(int)

scaler_standard = StandardScaler()
numeric_features = ['accountlength', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
                    'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
                    'totalnightcharge', 'totalintlminutes', 'totalintlcharge']  

scaler_minmax = MinMaxScaler()
numeric_features_minmax = ['numbervmailmessages', 'totalintlcalls', 'numbercustomerservicecalls']  

X_numeric_train_balanced = X_train_balanced[numeric_features] 
X_numeric_train_balanced_scaled = scaler_standard.fit_transform(X_numeric_train_balanced)  
X_train_balanced_scaled = X_train_balanced.copy()
X_train_balanced_scaled[numeric_features] = X_numeric_train_balanced_scaled  

X_numeric_minmax = X_train_balanced_scaled[numeric_features_minmax]  
X_numeric_minmax_scaled = scaler_minmax.fit_transform(X_numeric_minmax)  
X_train_balanced_scaled[numeric_features_minmax] = X_numeric_minmax_scaled

# Linear Kernel

svm_model_lb = SVC(kernel='linear', C=100)
svm_model_lb.fit(X_train_balanced_scaled, y_train_balanced)

train_preds = svm_model_lb.predict(X_train_balanced_scaled)
X_train_balanced_accuracy = accuracy_score(y_train_balanced, train_preds)
print(f"Accuracy on X_train_balanced data: {X_train_balanced_accuracy:.2f}")

print("Classification Report for X_train_balanced data:")
print(classification_report(y_train_balanced, train_preds, zero_division=0))

correct_predictions = (train_preds == y_train_balanced)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

train_conf_matrix = confusion_matrix(y_train_balanced, train_preds)

plt.figure(figsize=(4, 4))
sns.heatmap(train_conf_matrix, annot=True, fmt='g', cmap='Blues', 
             square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of SMV Linear Kernel')
plt.show()

Accuracy on X_train_balanced data: 0.80
Classification Report for X_train_balanced data:
              precision    recall  f1-score   support

           0       0.82      0.76      0.79      1719
           1       0.78      0.84      0.80      1719

    accuracy                           0.80      3438
   macro avg       0.80      0.80      0.80      3438
weighted avg       0.80      0.80      0.80      3438

Number of correct predictions: 2741

# RBF Kernel
svm_model_rbf_b = SVC(kernel='rbf', C=100)
svm_model_rbf_b.fit(X_train_balanced_scaled, y_train_balanced)

train_preds = svm_model_rbf_b.predict(X_train_balanced_scaled)
X_train_balanced_accuracy = accuracy_score(y_train_balanced, train_preds)
print(f"Accuracy on X_train_balanced data: {X_train_balanced_accuracy:.2f}")

print("Classification Report for X_train_balanced data:")
print(classification_report(y_train_balanced, train_preds, zero_division=0))

correct_predictions = (train_preds == y_train_balanced)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

train_conf_matrix = confusion_matrix(y_train_balanced, train_preds)

plt.figure(figsize=(4, 4))
sns.heatmap(train_conf_matrix, annot=True, fmt='g', cmap='Blues', 
             square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of SMV RBF Kernel')
plt.show()

Accuracy on X_train_balanced data: 1.00
Classification Report for X_train_balanced data:
              precision    recall  f1-score   support

           0       1.00      1.00      1.00      1719
           1       1.00      1.00      1.00      1719

    accuracy                           1.00      3438
   macro avg       1.00      1.00      1.00      3438
weighted avg       1.00      1.00      1.00      3438

Number of correct predictions: 3426

cv = KFold(n_splits=5, shuffle=True, random_state=42)

svm_model_rbf_b = SVC(kernel='rbf', C=100)
scores_rbf_b = cross_val_score(svm_model_rbf_b, X_train_balanced_scaled, y_train_balanced, cv=cv, scoring='accuracy')

print("Cross-validation scores for RBF Kernel:", scores_rbf_b)
print("Mean accuracy for RBF Kernel:", np.mean(scores_rbf_b))

svm_model_lb = SVC(kernel='linear', C=100)
scores_lb = cross_val_score(svm_model_lb, X_train_balanced_scaled, y_train_balanced, cv=cv, scoring='accuracy')

print("Cross-validation scores for Linear Kernel:", scores_lb)
print("Mean accuracy for Linear Kernel:", np.mean(scores_lb))

Cross-validation scores for RBF Kernel: [0.90988372 0.8997093  0.90261628 0.89665211 0.88500728]
Mean accuracy for RBF Kernel: 0.898773738194374
Cross-validation scores for Linear Kernel: [0.80377907 0.79215116 0.77180233 0.80494905 0.79039301]
Mean accuracy for Linear Kernel: 0.7926149250194644

clf = MLPClassifier().fit(X_train_balanced, y_train_balanced)

y_pred = clf.predict(X_test)

accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)
    
print("Accuracy:", accuracy)
print("Precision:", precision)
print("Recall:", recall)
print("F1 Score:", f1)

Accuracy: 0.679
Precision: 0.28743961352657005
Recall: 0.8206896551724138
F1 Score: 0.42576028622540246

mlp_gs = MLPClassifier(max_iter=100)
parameter_space = {
    'hidden_layer_sizes': [(50,), (100,), (200,), (10, 20), (10, 50, 10)],
    'activation': ['tanh', 'relu'],
    'solver': ['sgd', 'adam'],
    'alpha': [0.0001, 0.05],
    'learning_rate': ['constant','adaptive'],
}

clf = GridSearchCV(mlp_gs, parameter_space, n_jobs=-1, cv=5)
clf.fit(X_train_balanced, y_train_balanced)

C:\Users\Manuela\AppData\Local\Programs\Python\Python311\Lib\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (100) reached and the optimization hasn't converged yet.
  warnings.warn(

GridSearchCV(cv=5, estimator=MLPClassifier(max_iter=100), n_jobs=-1,
             param_grid={'activation': ['tanh', 'relu'],
                         'alpha': [0.0001, 0.05],
                         'hidden_layer_sizes': [(50,), (100,), (200,), (10, 20),
                                                (10, 50, 10)],
                         'learning_rate': ['constant', 'adaptive'],
                         'solver': ['sgd', 'adam']})

GridSearchCV(cv=5, estimator=MLPClassifier(max_iter=100), n_jobs=-1,
             param_grid={'activation': ['tanh', 'relu'],
                         'alpha': [0.0001, 0.05],
                         'hidden_layer_sizes': [(50,), (100,), (200,), (10, 20),
                                                (10, 50, 10)],
                         'learning_rate': ['constant', 'adaptive'],
                         'solver': ['sgd', 'adam']})

MLPClassifier(max_iter=100)

MLPClassifier(max_iter=100)

print('Best parameters found:\n', clf.best_params_)

Best parameters found:
 {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}

means = clf.cv_results_['mean_test_score']
stds = clf.cv_results_['std_test_score']
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
    print("%0.3f (+/-%0.03f) for %r" % (mean, std * 2, params))

0.639 (+/-0.031) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.771 (+/-0.022) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.652 (+/-0.040) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.763 (+/-0.024) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.668 (+/-0.029) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.784 (+/-0.054) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.686 (+/-0.062) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.785 (+/-0.045) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.689 (+/-0.040) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.807 (+/-0.056) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.708 (+/-0.033) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.812 (+/-0.058) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.603 (+/-0.069) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.619 (+/-0.030) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.609 (+/-0.116) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.617 (+/-0.125) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.604 (+/-0.052) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.621 (+/-0.043) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.609 (+/-0.028) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.646 (+/-0.083) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.646 (+/-0.053) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.764 (+/-0.051) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.641 (+/-0.043) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.749 (+/-0.052) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.667 (+/-0.032) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.794 (+/-0.038) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.667 (+/-0.043) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.807 (+/-0.040) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.698 (+/-0.037) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.816 (+/-0.049) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.697 (+/-0.048) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.817 (+/-0.050) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.538 (+/-0.066) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.678 (+/-0.123) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.593 (+/-0.096) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.660 (+/-0.102) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.619 (+/-0.067) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.651 (+/-0.035) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.614 (+/-0.088) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.600 (+/-0.082) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.669 (+/-0.028) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.784 (+/-0.057) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.654 (+/-0.040) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.771 (+/-0.071) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.672 (+/-0.012) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.775 (+/-0.074) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.676 (+/-0.032) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.772 (+/-0.105) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.691 (+/-0.042) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.805 (+/-0.053) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.692 (+/-0.042) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.788 (+/-0.068) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.628 (+/-0.127) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.645 (+/-0.079) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.572 (+/-0.138) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.715 (+/-0.076) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.627 (+/-0.131) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.691 (+/-0.063) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.621 (+/-0.119) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.629 (+/-0.148) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.650 (+/-0.060) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.769 (+/-0.034) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.665 (+/-0.030) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.766 (+/-0.028) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.671 (+/-0.052) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.780 (+/-0.064) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.675 (+/-0.053) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.776 (+/-0.056) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.691 (+/-0.041) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.808 (+/-0.057) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.696 (+/-0.051) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.783 (+/-0.048) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.631 (+/-0.082) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.700 (+/-0.097) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.543 (+/-0.122) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.685 (+/-0.099) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.600 (+/-0.154) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.721 (+/-0.060) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.656 (+/-0.041) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.702 (+/-0.125) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}

clf_b = MLPClassifier(activation= 'tanh', alpha = 0.05, hidden_layer_sizes = (200,), learning_rate = 'adaptive', solver = 'adam').fit(X_train_balanced, y_train_balanced)


y_pred = clf_b.predict(X_test)

accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)
    
print("Accuracy:", accuracy)
print("Precision:", precision)
print("Recall:", recall)
print("F1 Score:", f1)

Accuracy: 0.795
Precision: 0.38549618320610685
Recall: 0.696551724137931
F1 Score: 0.4963144963144963

print("Classification Report for X_train_balanced data:")
print(classification_report(y_test, y_pred, zero_division=0))

correct_predictions = (y_pred == y_test)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

Classification Report for X_train_balanced data:
              precision    recall  f1-score   support

         0.0       0.94      0.81      0.87       855
         1.0       0.39      0.70      0.50       145

    accuracy                           0.80      1000
   macro avg       0.66      0.75      0.68      1000
weighted avg       0.86      0.80      0.82      1000

Number of correct predictions: 795

NNC_conf_matrix = confusion_matrix(y_test, y_pred)

plt.figure(figsize=(4, 4))
sns.heatmap(KNN_conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of NNC')
plt.show()

scaler = StandardScaler()
X_train_scaler = scaler.fit_transform(X_train)
X_test_scaler = scaler.transform(X_test)

knn = KNeighborsClassifier(n_neighbors=2)
knn.fit(X_train_scaler, y_train)

KNeighborsClassifier(n_neighbors=2)

KNeighborsClassifier(n_neighbors=2)

y_pred = knn.predict(X_test_scaler)

accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)

Accuracy: 0.88

k_values = [i for i in range (1,30)]
scores = []

scaler = StandardScaler()
X_train = scaler.fit_transform(X_train_scaler)

for k in k_values:
    knn = KNeighborsClassifier(n_neighbors=k)
    knn.fit(X_train_scaler, y_train)
    y_pred = knn.predict(X_test_scaler)
    score = accuracy_score(y_test, y_pred)
    scores.append(np.mean(score))

sns.lineplot(x = k_values, y = scores, marker = 'o')
plt.xlabel("K Values")
plt.ylabel("Accuracy Score")

Text(0, 0.5, 'Accuracy Score')

knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(X_train_scaler, y_train)

y_pred = knn.predict(X_test_scaler)

accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)

Accuracy: 0.898

knn = KNeighborsClassifier(n_neighbors=3)
knn.fit(X_train_scaler, y_train)

y_pred = knn.predict(X_test_scaler)

accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)
    
print("Accuracy:", accuracy)
print("Precision:", precision)
print("Recall:", recall)
print("F1 Score:", f1)

Accuracy: 0.898
Precision: 0.7721518987341772
Recall: 0.4206896551724138
F1 Score: 0.5446428571428571

KNN_conf_matrix = confusion_matrix(y_test, y_pred)

plt.figure(figsize=(4, 4))
sns.heatmap(KNN_conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of KNN')
plt.show()

X_train = train_data.loc[:, train_data.columns != 'churn']
y_train = train_data['churn'].astype(int)

X_test = test_data.loc[:, test_data.columns != 'churn']
y_test = test_data['churn'].astype(int)

scaler_standard = StandardScaler()
numeric_features = ['accountlength', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
                    'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
                    'totalnightcharge', 'totalintlminutes', 'totalintlcharge']  

scaler_minmax = MinMaxScaler()
numeric_features_minmax = ['numbervmailmessages', 'totalintlcalls', 'numbercustomerservicecalls']  

X_numeric_train = X_train[numeric_features] 
X_numeric_train_scaled = scaler_standard.fit_transform(X_numeric_train)  
X_train_scaled = X_train.copy()
X_train_scaled[numeric_features] = X_numeric_train_scaled  

X_numeric_minmax = X_train_scaled[numeric_features_minmax]  
X_numeric_minmax_scaled = scaler_minmax.fit_transform(X_numeric_minmax)  
X_train_scaled[numeric_features_minmax] = X_numeric_minmax_scaled  

X_numeric_test = X_test[numeric_features] 
X_numeric_test_scaled = scaler_standard.transform(X_numeric_test)  
X_test_scaled = X_test.copy()
X_test_scaled[numeric_features] = X_numeric_test_scaled  

X_numeric_test_minmax = X_test_scaled[numeric_features_minmax]  
X_numeric_test_minmax_scaled = scaler_minmax.transform(X_numeric_test_minmax)  
X_test_scaled[numeric_features_minmax] = X_numeric_test_minmax_scaled  

model_NB = MixedNB()

model_NB.fit(X_train_scaled, y_train)  

preds = model_NB.predict(X_train_scaled)

accuracy = accuracy_score(y_train, preds)
print(f"Accuracy on train data: {accuracy:.2f}")

print("Classification Report for train data:")
print(classification_report(y_train, preds, zero_division=0))

correct_predictions = (preds == y_train)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

conf_matrix = confusion_matrix(y_train, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.show()

preds = model_NB.predict(X_test_scaled)

accuracy = accuracy_score(y_test, preds)
print(f"Accuracy on test data: {accuracy:.2f}")

print("Classification Report for test data:")
print(classification_report(y_test, preds, zero_division=0))

correct_predictions = (preds == y_test)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

conf_matrix = confusion_matrix(y_test, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of NB')
plt.show()

Accuracy on train data: 0.88
Classification Report for train data:
              precision    recall  f1-score   support

           0       0.92      0.94      0.93      3438
           1       0.58      0.53      0.55       562

    accuracy                           0.88      4000
   macro avg       0.75      0.73      0.74      4000
weighted avg       0.88      0.88      0.88      4000

Number of correct predictions: 3520

Accuracy on test data: 0.87
Classification Report for test data:
              precision    recall  f1-score   support

           0       0.92      0.93      0.92       855
           1       0.56      0.52      0.54       145

    accuracy                           0.87      1000
   macro avg       0.74      0.73      0.73      1000
weighted avg       0.87      0.87      0.87      1000

Number of correct predictions: 871

model = DecisionTreeClassifier(random_state = 0)

param_grid = {
    'max_depth': [None,4,8,14,16,20,22],
    'min_samples_split': [15,20,25,30,],
    'ccp_alpha':[0.0001,0.001,0.01]
}

grid_search = GridSearchCV(model, param_grid, cv=5)
grid_search.fit(X_train, y_train)

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 4, 8, 14, 16, 20, 22],
                         'min_samples_split': [15, 20, 25, 30]})

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 4, 8, 14, 16, 20, 22],
                         'min_samples_split': [15, 20, 25, 30]})

DecisionTreeClassifier(random_state=0)

DecisionTreeClassifier(random_state=0)

best_dt_model = grid_search.best_estimator_
best_dt_model

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, min_samples_split=25,
                       random_state=0)

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, min_samples_split=25,
                       random_state=0)

plt.subplots(nrows = 1,ncols = 1,figsize = (14,14), dpi=200)
tree.plot_tree(best_dt_model, fontsize=5, filled=True)
plt.title('Decision Tree with the best parameters defined from a Cross Validation Method')
plt.show()

def scores(model):
    preds = model.predict(X_test)
    print("Classification Report for test data:")
    print(classification_report(y_test, preds, zero_division=0))
scores(best_dt_model)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.95      0.99      0.97       855
         1.0       0.90      0.67      0.77       145

    accuracy                           0.94      1000
   macro avg       0.92      0.83      0.87      1000
weighted avg       0.94      0.94      0.94      1000

preds = best_dt_model.predict(X_test)
cm = confusion_matrix(y_test, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', cbar=True)
plt.xlabel('Predicted Labels')
plt.ylabel('True Labels')
plt.title('Confusion Matrix of Decision Tree Model')
plt.show()

model_f1 = DecisionTreeClassifier(random_state = 0)

grid_search_f1 = GridSearchCV(model_f1, param_grid, scoring='f1', cv=5)
grid_search_f1.fit(X_train, y_train)

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 4, 8, 14, 16, 20, 22],
                         'min_samples_split': [15, 20, 25, 30]},
             scoring='f1')

GridSearchCV(cv=5, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'ccp_alpha': [0.0001, 0.001, 0.01],
                         'max_depth': [None, 4, 8, 14, 16, 20, 22],
                         'min_samples_split': [15, 20, 25, 30]},
             scoring='f1')

DecisionTreeClassifier(random_state=0)

DecisionTreeClassifier(random_state=0)

best_dt_model_f1 = grid_search_f1.best_estimator_

best_dt_model_f1

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, min_samples_split=25,
                       random_state=0)

DecisionTreeClassifier(ccp_alpha=0.001, max_depth=16, min_samples_split=25,
                       random_state=0)

scores(best_dt_model_f1)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.95      0.99      0.97       855
         1.0       0.90      0.67      0.77       145

    accuracy                           0.94      1000
   macro avg       0.92      0.83      0.87      1000
weighted avg       0.94      0.94      0.94      1000

preds = best_dt_model_f1.predict(X_test)
cm = confusion_matrix(y_test, preds)

plt.figure(figsize=(4, 4))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', cbar=True)
plt.xlabel('Predicted Labels')
plt.ylabel('True Labels')
plt.title('Confusion Matrix of the F1 Tunned Decision Tree Model')
plt.show()

X_train_tune, X_train_final, y_train_tune, y_train_final = train_test_split(X_train, y_train, test_size=0.8, random_state=0)

bag = BaggingClassifier(random_state=0)

param_grid = {
    'n_estimators': [100, 150, 200, 250]
}

grid_search = GridSearchCV(bag, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_bag = grid_search.best_estimator_

best_bag

BaggingClassifier(n_estimators=200, random_state=0)

BaggingClassifier(n_estimators=200, random_state=0)

base_tree = DecisionTreeClassifier()
bag = BaggingClassifier(n_estimators=200, random_state=0)

bag.fit(X_train, y_train)
scores(bag)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.96      0.99      0.97       855
         1.0       0.90      0.73      0.81       145

    accuracy                           0.95      1000
   macro avg       0.93      0.86      0.89      1000
weighted avg       0.95      0.95      0.95      1000

rf = RandomForestClassifier(random_state=0)
param_grid = {
    'n_estimators': [100, 150, 200, 250],
    'max_features': [None, 'sqrt', 'log2']  
}

grid_search = GridSearchCV(rf, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_rf = grid_search.best_estimator_

best_rf

RandomForestClassifier(max_features=None, n_estimators=250, random_state=0)

RandomForestClassifier(max_features=None, n_estimators=250, random_state=0)

rf = RandomForestClassifier(n_estimators=200, random_state=0)
rf.fit(X_train,y_train)
scores(rf)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.96      0.99      0.97       855
         1.0       0.91      0.75      0.82       145

    accuracy                           0.95      1000
   macro avg       0.93      0.87      0.90      1000
weighted avg       0.95      0.95      0.95      1000

model =  AdaBoostClassifier(random_state=0)
param_grid = {
    'n_estimators': [100, 150, 200, 250]
}

grid_search = GridSearchCV(model, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_ada = grid_search.best_estimator_
best_ada

AdaBoostClassifier(n_estimators=100, random_state=0)

AdaBoostClassifier(n_estimators=100, random_state=0)

adaboosted_tree = AdaBoostClassifier(n_estimators=150, random_state=0)
adaboosted_tree.fit(X_train, y_train)
scores(adaboosted_tree)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.90      0.96      0.93       855
         1.0       0.61      0.37      0.46       145

    accuracy                           0.87      1000
   macro avg       0.75      0.66      0.69      1000
weighted avg       0.86      0.87      0.86      1000

model =  GBC(random_state=0)
param_grid = {
    'n_estimators': [100, 150, 200, 250]
}

grid_search = GridSearchCV(model, param_grid, cv=3, scoring='accuracy')
grid_search.fit(X_train_tune, y_train_tune)
best_gbc = grid_search.best_estimator_
best_params = grid_search.best_params_
best_params

{'n_estimators': 150}

gbc_boosted_tree = GBC(n_estimators=100, random_state=0)
gbc_boosted_tree.fit(X_train, y_train)
scores(gbc_boosted_tree)

Classification Report for test data:
              precision    recall  f1-score   support

         0.0       0.96      0.99      0.97       855
         1.0       0.91      0.75      0.82       145

    accuracy                           0.95      1000
   macro avg       0.93      0.87      0.90      1000
weighted avg       0.95      0.95      0.95      1000

X_train = train_data.loc[:, train_data.columns != 'churn']
y_train = train_data['churn'].astype(int)

X_test = test_data.loc[:, test_data.columns != 'churn']
y_test = test_data['churn'].astype(int)

scaler_standard = StandardScaler()
numeric_features = ['accountlength', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
                    'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
                    'totalnightcharge', 'totalintlminutes', 'totalintlcharge']  

scaler_minmax = MinMaxScaler()
numeric_features_minmax = ['numbervmailmessages', 'totalintlcalls', 'numbercustomerservicecalls']  

X_numeric_train = X_train[numeric_features] 
X_numeric_train_scaled = scaler_standard.fit_transform(X_numeric_train)  
X_train_scaled = X_train.copy()
X_train_scaled[numeric_features] = X_numeric_train_scaled  

X_numeric_minmax = X_train_scaled[numeric_features_minmax]  
X_numeric_minmax_scaled = scaler_minmax.fit_transform(X_numeric_minmax)  
X_train_scaled[numeric_features_minmax] = X_numeric_minmax_scaled  

X_numeric_test = X_test[numeric_features] 
X_numeric_test_scaled = scaler_standard.transform(X_numeric_test)  
X_test_scaled = X_test.copy()
X_test_scaled[numeric_features] = X_numeric_test_scaled  

X_numeric_test_minmax = X_test_scaled[numeric_features_minmax]  
X_numeric_test_minmax_scaled = scaler_minmax.transform(X_numeric_test_minmax)  
X_test_scaled[numeric_features_minmax] = X_numeric_test_minmax_scaled

# Linear Kernel

svm_model_l = SVC(kernel='linear', C=100)
svm_model_l.fit(X_train_scaled, y_train)

train_preds = svm_model_l.predict(X_train_scaled)
train_accuracy = accuracy_score(y_train, train_preds)
print(f"Accuracy on train data: {train_accuracy:.2f}")

print("Classification Report for train data:")
print(classification_report(y_train, train_preds, zero_division=0))

correct_predictions = (train_preds == y_train)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

train_conf_matrix = confusion_matrix(y_train, train_preds)

plt.figure(figsize=(4, 4))
sns.heatmap(train_conf_matrix, annot=True, fmt='g', cmap='Blues', 
             square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.show()

test_preds = svm_model_l.predict(X_test_scaled)
test_accuracy = accuracy_score(y_test, test_preds)
print(f"Accuracy on test data: {test_accuracy:.2f}")

print("Classification Report for test data:")
print(classification_report(y_test, test_preds, zero_division=0))

correct_predictions_test = (test_preds == y_test)
num_correct_test = sum(correct_predictions_test)
print(f"Number of correct predictions on test data: {num_correct_test}")

test_conf_matrix = confusion_matrix(y_test, test_preds)

plt.figure(figsize=(4, 4))
sns.heatmap(test_conf_matrix, annot=True, fmt='g', cmap='Blues', 
             square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of SMV Linear Kernel')
plt.show()

Accuracy on train data: 0.86
Classification Report for train data:
              precision    recall  f1-score   support

           0       0.86      1.00      0.92      3438
           1       0.00      0.00      0.00       562

    accuracy                           0.86      4000
   macro avg       0.43      0.50      0.46      4000
weighted avg       0.74      0.86      0.79      4000

Number of correct predictions: 3438

Accuracy on test data: 0.85
Classification Report for test data:
              precision    recall  f1-score   support

           0       0.85      1.00      0.92       855
           1       0.00      0.00      0.00       145

    accuracy                           0.85      1000
   macro avg       0.43      0.50      0.46      1000
weighted avg       0.73      0.85      0.79      1000

Number of correct predictions on test data: 855

svm_model_rbf = SVC(kernel='rbf', C=100)
svm_model_rbf.fit(X_train_scaled, y_train)

train_preds = svm_model_rbf.predict(X_train_scaled)
train_accuracy = accuracy_score(y_train, train_preds)
print(f"Accuracy on train data: {train_accuracy:.2f}")

print("Classification Report for train data:")
print(classification_report(y_train, train_preds, zero_division=0))

correct_predictions = (train_preds == y_train)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

train_conf_matrix = confusion_matrix(y_train, train_preds)

plt.figure(figsize=(4, 4))
sns.heatmap(train_conf_matrix, annot=True, fmt='g', cmap='Blues', 
             square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.show()

test_preds = svm_model_rbf.predict(X_test_scaled)
test_accuracy = accuracy_score(y_test, test_preds)
print(f"Accuracy on test data: {test_accuracy:.2f}")

print("Classification Report for test data:")
print(classification_report(y_test, test_preds, zero_division=0))

correct_predictions_test = (test_preds == y_test)
num_correct_test = sum(correct_predictions_test)
print(f"Number of correct predictions on test data: {num_correct_test}")

test_conf_matrix = confusion_matrix(y_test, test_preds)

plt.figure(figsize=(4, 4))
sns.heatmap(test_conf_matrix, annot=True, fmt='g', cmap='Blues', 
             square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of SMV RBF Kernel')
plt.show()

Accuracy on train data: 0.99
Classification Report for train data:
              precision    recall  f1-score   support

           0       0.99      1.00      1.00      3438
           1       1.00      0.94      0.97       562

    accuracy                           0.99      4000
   macro avg       0.99      0.97      0.98      4000
weighted avg       0.99      0.99      0.99      4000

Number of correct predictions: 3967

Accuracy on test data: 0.88
Classification Report for test data:
              precision    recall  f1-score   support

           0       0.93      0.93      0.93       855
           1       0.59      0.57      0.58       145

    accuracy                           0.88      1000
   macro avg       0.76      0.75      0.75      1000
weighted avg       0.88      0.88      0.88      1000

Number of correct predictions on test data: 880

clf = MLPClassifier().fit(X_train, y_train)

y_pred = clf.predict(X_test)

accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)
    
print("Accuracy:", accuracy)
print("Precision:", precision)
print("Recall:", recall)
print("F1 Score:", f1)

Accuracy: 0.873
Precision: 0.6730769230769231
Recall: 0.2413793103448276
F1 Score: 0.3553299492385787

mlp_gs = MLPClassifier(max_iter=100)
parameter_space = {
    'hidden_layer_sizes': [(50,), (100,), (200,), (10, 20), (10, 50, 10)],
    'activation': ['tanh', 'relu'],
    'solver': ['sgd', 'adam'],
    'alpha': [0.0001, 0.05],
    'learning_rate': ['constant','adaptive'],
}
from sklearn.model_selection import GridSearchCV
clf = GridSearchCV(mlp_gs, parameter_space, n_jobs=-1, cv=5)
clf.fit(X_train, y_train)

GridSearchCV(cv=5, estimator=MLPClassifier(max_iter=100), n_jobs=-1,
             param_grid={'activation': ['tanh', 'relu'],
                         'alpha': [0.0001, 0.05],
                         'hidden_layer_sizes': [(50,), (100,), (200,), (10, 20),
                                                (10, 50, 10)],
                         'learning_rate': ['constant', 'adaptive'],
                         'solver': ['sgd', 'adam']})

GridSearchCV(cv=5, estimator=MLPClassifier(max_iter=100), n_jobs=-1,
             param_grid={'activation': ['tanh', 'relu'],
                         'alpha': [0.0001, 0.05],
                         'hidden_layer_sizes': [(50,), (100,), (200,), (10, 20),
                                                (10, 50, 10)],
                         'learning_rate': ['constant', 'adaptive'],
                         'solver': ['sgd', 'adam']})

MLPClassifier(max_iter=100)

MLPClassifier(max_iter=100)

print('Best parameters found:\n', clf.best_params_)

Best parameters found:
 {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}

means = clf.cv_results_['mean_test_score']
stds = clf.cv_results_['std_test_score']
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
    print("%0.3f (+/-%0.03f) for %r" % (mean, std * 2, params))

0.860 (+/-0.003) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.862 (+/-0.009) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.861 (+/-0.009) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.866 (+/-0.006) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.862 (+/-0.005) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.865 (+/-0.006) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.860 (+/-0.005) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.868 (+/-0.011) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.869 (+/-0.011) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.868 (+/-0.006) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.869 (+/-0.005) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.871 (+/-0.012) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.860 (+/-0.002) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.860 (+/-0.002) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.860 (+/-0.002) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.861 (+/-0.010) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.860 (+/-0.001) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.860 (+/-0.003) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.859 (+/-0.002) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.861 (+/-0.012) for {'activation': 'tanh', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.861 (+/-0.005) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.866 (+/-0.010) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.861 (+/-0.010) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.867 (+/-0.007) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.866 (+/-0.006) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.868 (+/-0.009) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.863 (+/-0.011) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.868 (+/-0.019) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.866 (+/-0.003) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.864 (+/-0.005) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.869 (+/-0.008) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.869 (+/-0.012) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.860 (+/-0.001) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.860 (+/-0.005) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.859 (+/-0.001) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.860 (+/-0.001) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.860 (+/-0.003) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.862 (+/-0.006) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.860 (+/-0.001) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.861 (+/-0.008) for {'activation': 'tanh', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.865 (+/-0.010) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.856 (+/-0.017) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.863 (+/-0.008) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.864 (+/-0.014) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.864 (+/-0.010) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.863 (+/-0.007) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.866 (+/-0.007) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.868 (+/-0.008) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.868 (+/-0.015) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.857 (+/-0.042) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.869 (+/-0.014) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.867 (+/-0.012) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.860 (+/-0.001) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.865 (+/-0.010) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.859 (+/-0.001) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.867 (+/-0.005) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.862 (+/-0.005) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.871 (+/-0.008) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.860 (+/-0.002) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.862 (+/-0.015) for {'activation': 'relu', 'alpha': 0.0001, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.867 (+/-0.010) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.854 (+/-0.018) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'constant', 'solver': 'adam'}
0.859 (+/-0.005) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.858 (+/-0.028) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.869 (+/-0.009) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.863 (+/-0.025) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'constant', 'solver': 'adam'}
0.867 (+/-0.008) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.855 (+/-0.032) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.865 (+/-0.015) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'sgd'}
0.866 (+/-0.015) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'constant', 'solver': 'adam'}
0.869 (+/-0.010) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.865 (+/-0.010) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (200,), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.862 (+/-0.011) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'sgd'}
0.862 (+/-0.008) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'constant', 'solver': 'adam'}
0.863 (+/-0.015) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.865 (+/-0.013) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 20), 'learning_rate': 'adaptive', 'solver': 'adam'}
0.865 (+/-0.011) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'sgd'}
0.866 (+/-0.004) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'constant', 'solver': 'adam'}
0.862 (+/-0.005) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'sgd'}
0.865 (+/-0.011) for {'activation': 'relu', 'alpha': 0.05, 'hidden_layer_sizes': (10, 50, 10), 'learning_rate': 'adaptive', 'solver': 'adam'}

clf = MLPClassifier(activation= 'relu', alpha = 0.0001, hidden_layer_sizes = (10, 50, 10), learning_rate = 'constant', solver = 'adam').fit(X_train, y_train)


y_pred = clf_b.predict(X_test)

accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)
    
print("Accuracy:", accuracy)
print("Precision:", precision)
print("Recall:", recall)
print("F1 Score:", f1)

Accuracy: 0.795
Precision: 0.38549618320610685
Recall: 0.696551724137931
F1 Score: 0.4963144963144963

print("Classification Report for X_train data:")
print(classification_report(y_test, y_pred, zero_division=0))

correct_predictions = (y_pred == y_test)
num_correct = sum(correct_predictions)
print(f"Number of correct predictions: {num_correct}")

Classification Report for X_train data:
              precision    recall  f1-score   support

           0       0.94      0.81      0.87       855
           1       0.39      0.70      0.50       145

    accuracy                           0.80      1000
   macro avg       0.66      0.75      0.68      1000
weighted avg       0.86      0.80      0.82      1000

Number of correct predictions: 795

NNC_conf_matrix = confusion_matrix(y_test, y_pred)

plt.figure(figsize=(4, 4))
sns.heatmap(NNC_conf_matrix, annot=True, fmt='g', cmap='Blues', 
            square=True,
            xticklabels=['No', 'Yes'], 
            yticklabels=['No', 'Yes'])
plt.xlabel('Predicted labels')
plt.ylabel('True labels')
plt.title('Confusion Matrix of NNC')
plt.show()

models = [best_dt_model_f1_b, best_dt_model_f1, best_bag_b, bag, best_rf_b, rf, best_ada_b, adaboosted_tree, best_gbc_b, gbc_boosted_tree]
models_names = ['Decision Tree', 'Decision Tree', 'Bagging', 'Bagging', 'Random Forest', 'Random Forest', 'Ada Booster', 'Ada Booster', 'Gradient Boosting', 'Gradient Boosting']
data = ["Balanced", "Unbalanced", "Balanced", "Unbalanced", "Balanced", "Unbalanced", "Balanced", "Unbalanced", "Balanced", "Unbalanced"]
accuracies = []
recall_scores = []
precision_scores = []
f1_scores = []

for model in models:
    y_pred = model.predict(X_test)
    accuracy = accuracy_score(y_test, y_pred)
    recall = recall_score(y_test, y_pred)
    precision = precision_score(y_test, y_pred)
    f1 = f1_score(y_test, y_pred)
    accuracies.append(accuracy)
    recall_scores.append(recall)
    precision_scores.append(precision)
    f1_scores.append(f1)

comp = {'Model': models_names, 'Data': data, 'Accuracy': accuracies, 'Recall Score': recall_scores, 'Precision Score': precision_scores, 'F1 Score': f1_scores}
comparation = pd.DataFrame(comp)

comparation

import pandas as pd

models = ['knn_b', 'knn', 'clf_b', 'clf', 'model_NB_b', 'model_NB', 'svm_model_lb', 'svm_model_l']
models_names = ['KNN', 'KNN', 'NeuralNetwork', 'NeuralNetwork', 'Bayes', 'Bayes','SVMLinear', 'SVMLinear']
data = ["Balanced", "Unbalanced", "Balanced", "Unbalanced","Balanced", "Unbalanced", "Balanced", "Unbalanced"]
accuracies = []
recall_scores = []
precision_scores = []
f1_scores = []

for model in models:   
    if model == 'knn_b':
        accuracies.append(0.80)
        recall_scores.append(0.80)
        precision_scores.append(0.87)
        f1_scores.append(0.82)
    if model == 'knn':
        accuracies.append(0.89)
        recall_scores.append(0.42)
        precision_scores.append(0.77)
        f1_scores.append(0.54)
    if model == 'clf_b':
        accuracies.append(0.80)
        recall_scores.append(0.86)
        precision_scores.append(0.80)
        f1_scores.append(0.82)
    if model == 'clf':
        accuracies.append(0.80)
        recall_scores.append(0.80)
        precision_scores.append(0.86)
        f1_scores.append(0.82)
    if model == 'model_NB_b':
        accuracies.append(0.78)
        recall_scores.append(0.79)
        precision_scores.append(0.78)
        f1_scores.append(0.78)
    elif model == 'model_NB':
        accuracies.append(0.87)
        recall_scores.append(0.87)
        precision_scores.append(0.87 )
        f1_scores.append(0.87)
    elif model == 'svm_model_lb':
        accuracies.append(0.80)
        recall_scores.append(0.80)
        precision_scores.append(0.80)
        f1_scores.append(0.80)
    elif model == "svm_model_l":
        accuracies.append(0.85)
        recall_scores.append(0.85)
        precision_scores.append(0.73)
        f1_scores.append(0.79)

comp = {'Model': models_names, 'Data': data, 'Accuracy': accuracies, 'Recall Score': recall_scores, 'Precision Score': precision_scores, 'F1 Score': f1_scores}
comparison = pd.DataFrame(comp)

comparison

voting_model_balanced = VotingClassifier(estimators=[
    ('best_bag_b', best_bag_b),
    ('best_rf_b', best_rf_b),
    ('best_gbc_b', best_gbc_b)
], voting='hard')

voting_model_balanced.fit(X_train_balanced, y_train_balanced)

y_pred_balanced = voting_model_balanced.predict(X_test)

accuracy_balanced = accuracy_score(y_test, y_pred_balanced)
recall_balanced = recall_score(y_test, y_pred_balanced)
precision_balanced = precision_score(y_test, y_pred_balanced)
f1_balanced = f1_score(y_test, y_pred_balanced)

voting_model_unbalanced = VotingClassifier(estimators=[
    ('bag', bag),
    ('rf', rf),
    ('gbc_boosted_tree', gbc_boosted_tree)
], voting='hard')

voting_model_unbalanced.fit(X_train, y_train)

y_pred_unbalanced = voting_model_unbalanced.predict(X_test)

accuracy_unbalanced = accuracy_score(y_test, y_pred_unbalanced)
recall_unbalanced = recall_score(y_test, y_pred_unbalanced)
precision_unbalanced = precision_score(y_test, y_pred_unbalanced)
f1_unbalanced = f1_score(y_test, y_pred_unbalanced)

voting_comp = {
    'Data': ['Balanced', 'Unbalanced'],
    'Accuracy': [accuracy_balanced, accuracy_unbalanced],
    'Recall Score': [recall_balanced, recall_unbalanced],
    'Precision Score': [precision_balanced, precision_unbalanced],
    'F1 Score': [f1_balanced, f1_unbalanced]
}

voting_comparison = pd.DataFrame(voting_comp)

display(voting_comparison)

models = [best_dt_model_f1_b, best_dt_model_f1, best_bag_b, bag, best_rf_b, rf, best_ada_b, adaboosted_tree, best_gbc_b, gbc_boosted_tree]
predictions = pd.DataFrame({'True_Labels': y_test}, index=X_test.index)

for model in models:
    model_name = str(model).split('(')[0]
    predictions[model_name] = model.predict(X_test)

# Create a DataFrame for misclassified cases with the correct index
misclassified_cases = pd.DataFrame({'True_Labels': y_test}, index=X_test.index)

for model in models:
    model_name = str(model).split('(')[0]
    misclassified_cases[model_name] = (predictions['True_Labels'] != predictions[model_name]).astype(int)

misclassified_cases['Total_Misclassifications'] = misclassified_cases.iloc[:, 1:].sum(axis=1)

most_missclassified_cases = misclassified_cases[misclassified_cases['Total_Misclassifications'] > 2]

# Use boolean indexing to filter rows in X_test
most_missclassified_features = X_test.loc[most_missclassified_cases.index]

miss_features = test_data.loc[most_missclassified_cases.index]
miss_features

miss_features.describe()

variables = ['churn', 'internationalplan', 'voicemailplan']
fig, ax = plt.subplots(1, 3, figsize=(15, 5))

titles = ['Churn', 'International Plan', 'Voicemail Plan']

for i, variable in enumerate(variables):
    miss_features[variable].value_counts().plot(kind='bar', ax=ax[i])
    ax[i].set_title(titles[i]) 

plt.tight_layout()

fig, ax = plt.subplots(3, 5, figsize=(15, 10)) 

variables = ['accountlength', 'numbervmailmessages', 'totaldayminutes', 'totaldaycalls', 'totaldaycharge',
             'totaleveminutes', 'totalevecalls', 'totalevecharge', 'totalnightminutes', 'totalnightcalls',
             'totalnightcharge', 'totalintlminutes', 'totalintlcalls', 'totalintlcharge', 'numbercustomerservicecalls']

ax = ax.flatten()

for i, variable in enumerate(variables):
    miss_features[variable].plot.density(ax=ax[i], legend=True)
    ax[i].set_title(variable)

plt.tight_layout()
plt.show()

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
churn	1.000000	0.017599	0.223803	-0.096899	-0.098410	0.204437	0.015782	0.205893	0.089790	-0.007707	0.089355	0.043279	-0.007284	0.045688	0.061712	-0.044919	0.059816	0.216391
accountlength	0.017599	1.000000	0.004607	-0.020104	-0.017260	-0.001634	0.026951	-0.000937	-0.007911	0.011359	-0.008081	0.003669	-0.014938	0.003219	0.000348	0.013088	-0.000795	-0.001303
internationalplan	0.223803	0.004607	1.000000	-0.002055	0.006188	0.022611	0.000579	0.023479	0.032190	-0.018647	0.032162	-0.007305	0.016493	-0.009109	0.022755	0.002776	0.021917	-0.015707
voicemailplan	-0.096899	-0.020104	-0.002055	1.000000	0.876245	-0.000694	0.006088	-0.000403	0.030899	-0.003010	0.030819	0.004525	0.010740	0.004073	-0.002994	-0.000448	-0.003718	-0.005681
numbervmailmessages	-0.098410	-0.017260	0.006188	0.876245	1.000000	0.007037	0.003585	0.007447	0.020998	-0.007010	0.021033	0.005377	0.000477	0.005103	0.005578	-0.004427	0.003519	-0.009512
totaldayminutes	0.204437	-0.001634	0.022611	-0.000694	0.007037	1.000000	0.002713	0.989143	-0.010498	0.008244	-0.011064	0.009606	0.004540	0.009887	-0.021864	0.002713	-0.017980	-0.004387
totaldaycalls	0.015782	0.026951	0.000579	0.006088	0.003585	0.002713	1.000000	0.000978	-0.000310	0.002941	0.000457	0.005971	-0.005723	0.004675	0.015280	0.011587	0.012926	-0.008329
totaldaycharge	0.205893	-0.000937	0.023479	-0.000403	0.007447	0.989143	0.000978	1.000000	-0.008848	0.008896	-0.009529	0.011760	0.005581	0.012025	-0.022457	0.000308	-0.018685	-0.000983
totaleveminutes	0.089790	-0.007911	0.032190	0.030899	0.020998	-0.010498	-0.000310	-0.008848	1.000000	0.004539	0.993895	-0.017636	0.013641	-0.017142	0.000512	0.010567	-0.005727	-0.012021
totalevecalls	-0.007707	0.011359	-0.018647	-0.003010	-0.007010	0.008244	0.002941	0.008896	0.004539	1.000000	0.003882	0.003286	-0.013501	0.002488	-0.006834	0.001308	-0.005892	0.005579
totalevecharge	0.089355	-0.008081	0.032162	0.030819	0.021033	-0.011064	0.000457	-0.009529	0.993895	0.003882	1.000000	-0.017165	0.011710	-0.016676	0.001482	0.010663	-0.004732	-0.012234
totalnightminutes	0.043279	0.003669	-0.007305	0.004525	0.005377	0.009606	0.005971	0.011760	-0.017636	0.003286	-0.017165	1.000000	0.025320	0.993036	-0.006254	-0.019283	-0.005600	-0.006955
totalnightcalls	-0.007284	-0.014938	0.016493	0.010740	0.000477	0.004540	-0.005723	0.005581	0.013641	-0.013501	0.011710	0.025320	1.000000	0.027563	0.000101	-0.001553	0.002806	-0.006382
totalnightcharge	0.045688	0.003219	-0.009109	0.004073	0.005103	0.009887	0.004675	0.012025	-0.017142	0.002488	-0.016676	0.993036	0.027563	1.000000	-0.007700	-0.018429	-0.007068	-0.008293
totalintlminutes	0.061712	0.000348	0.022755	-0.002994	0.005578	-0.021864	0.015280	-0.022457	0.000512	-0.006834	0.001482	-0.006254	0.000101	-0.007700	1.000000	0.018314	0.984733	-0.011243
totalintlcalls	-0.044919	0.013088	0.002776	-0.000448	-0.004427	0.002713	0.011587	0.000308	0.010567	0.001308	0.010663	-0.019283	-0.001553	-0.018429	0.018314	1.000000	0.021163	-0.018942
totalintlcharge	0.059816	-0.000795	0.021917	-0.003718	0.003519	-0.017980	0.012926	-0.018685	-0.005727	-0.005892	-0.004732	-0.005600	0.002806	-0.007068	0.984733	0.021163	1.000000	-0.011175
numbercustomerservicecalls	0.216391	-0.001303	-0.015707	-0.005681	-0.009512	-0.004387	-0.008329	-0.000983	-0.012021	0.005579	-0.012234	-0.006955	-0.006382	-0.008293	-0.011243	-0.018942	-0.011175	1.000000

	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
3524	22.000000	0.000000	0.0	0.000000	222.900000	97.000000	37.890000	172.600000	91.000000	14.670000	187.600000	99.000000	8.440000	10.100000	3.000000	2.730000	1.000000
2478	172.000000	0.000000	0.0	0.000000	211.700000	100.000000	35.990000	198.700000	101.000000	16.890000	301.700000	136.000000	13.580000	6.500000	9.000000	1.760000	1.000000
717	96.000000	0.000000	0.0	0.000000	170.500000	86.000000	28.990000	277.500000	88.000000	23.590000	162.500000	117.000000	7.310000	12.200000	6.000000	3.290000	1.000000
3448	86.000000	0.000000	0.0	0.000000	178.300000	94.000000	30.310000	264.400000	74.000000	22.470000	284.600000	109.000000	12.810000	15.000000	2.000000	4.050000	0.000000
2234	127.000000	0.000000	0.0	0.000000	182.300000	124.000000	30.990000	169.900000	110.000000	14.440000	184.000000	116.000000	8.280000	9.300000	3.000000	2.510000	1.000000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
5152	144.564041	0.503050	1.0	36.012198	240.861576	94.435959	40.943474	220.167675	125.957306	18.717255	191.531715	72.478653	8.616433	7.055794	3.490851	1.906555	1.496950
5153	104.668934	0.000000	0.0	0.000000	162.843956	81.387636	27.681929	190.288161	79.775271	16.176003	143.768254	78.262442	6.473756	9.408748	3.000000	2.542624	5.912519
5154	78.135668	0.946957	0.0	0.000000	220.712955	77.007821	37.523043	276.288609	121.310439	23.480983	239.669738	112.371302	10.787426	11.230262	4.893914	3.033852	1.053043
5155	82.899100	0.000000	0.0	0.000000	288.970929	100.911090	49.127982	233.689910	97.341659	19.860889	223.773327	147.696303	10.072148	12.779720	2.189810	3.451638	1.924076
5156	87.024894	0.000000	0.0	0.000000	93.093592	102.256886	15.826372	212.238268	116.154662	18.041009	225.963824	120.281779	10.171782	8.556356	3.359110	2.310498	3.051112

	Model	Data	Accuracy	Recall Score	Precision Score	F1 Score
0	Decision Tree	Balanced	0.910	0.820690	0.650273	0.725610
1	Decision Tree	Unbalanced	0.941	0.668966	0.898148	0.766798
2	Bagging	Balanced	0.940	0.834483	0.770701	0.801325
3	Bagging	Unbalanced	0.949	0.731034	0.898305	0.806084
4	Random Forest	Balanced	0.938	0.827586	0.764331	0.794702
5	Random Forest	Unbalanced	0.953	0.751724	0.908333	0.822642
6	Ada Booster	Balanced	0.851	0.648276	0.489583	0.557864
7	Ada Booster	Unbalanced	0.874	0.365517	0.609195	0.456897
8	Gradient Boosting	Balanced	0.939	0.820690	0.772727	0.795987
9	Gradient Boosting	Unbalanced	0.953	0.751724	0.908333	0.822642

	Data	Accuracy	Recall Score	Precision Score	F1 Score
0	Balanced	0.944	0.827586	0.794702	0.810811
1	Unbalanced	0.956	0.751724	0.931624	0.832061

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
57	1.0	121.0	0.0	1.0	30.0	198.4	129.0	33.730000	75.30	77.0	6.40	181.2	77.0	8.150	5.80	3.0	1.570000	3.0
951	0.0	101.0	0.0	0.0	0.0	153.8	89.0	26.150000	234.00	89.0	19.89	196.3	77.0	8.830	11.60	2.0	3.130000	4.0
2369	1.0	112.0	0.0	0.0	0.0	174.3	123.0	29.630000	191.38	124.0	11.92	215.4	89.0	9.690	9.00	6.0	2.430000	4.0
3256	0.0	115.0	0.0	0.0	0.0	268.0	115.0	45.560000	153.60	106.0	13.06	232.3	65.0	10.450	17.00	1.0	4.590000	3.0
1165	0.0	50.0	1.0	1.0	26.0	307.1	94.0	52.210000	289.40	78.0	24.60	174.9	109.0	7.870	8.00	3.0	2.160000	0.0
2731	1.0	127.0	0.0	0.0	0.0	245.2	91.0	41.680000	217.20	92.0	18.46	243.1	128.0	10.940	13.90	6.0	3.750000	0.0
2745	1.0	61.0	0.0	1.0	40.0	105.0	78.0	17.850000	180.60	100.0	15.35	174.1	115.0	7.830	10.20	2.0	2.750000	2.0
3604	1.0	60.0	0.0	0.0	0.0	221.0	94.0	37.570000	308.20	72.0	26.20	233.2	94.0	10.490	10.10	5.0	2.730000	1.0
3070	1.0	154.0	0.0	0.0	0.0	154.5	122.0	26.270000	214.20	71.0	18.21	178.0	105.0	8.010	12.00	2.0	3.240000	3.0
437	1.0	100.0	0.0	0.0	0.0	278.0	76.0	47.260000	176.70	74.0	15.02	219.5	126.0	9.880	8.30	4.0	2.240000	0.0
1965	1.0	139.0	0.0	0.0	0.0	236.6	109.0	40.220000	169.90	107.0	14.44	212.3	118.0	9.550	11.10	2.0	3.000000	1.0
3873	1.0	63.0	0.0	0.0	0.0	162.5	86.0	27.630000	165.40	112.0	14.06	163.7	67.0	7.370	13.40	6.0	3.620000	4.0
2961	1.0	98.0	0.0	1.0	36.0	168.0	81.0	28.560000	163.20	125.0	13.87	172.7	120.0	7.770	8.00	2.0	2.160000	6.0
1551	1.0	225.0	0.0	0.0	0.0	165.4	106.0	28.120000	273.70	109.0	23.26	210.0	93.0	9.450	8.70	3.0	2.350000	0.0
2377	1.0	101.0	0.0	1.0	36.0	123.7	125.0	21.030000	172.60	106.0	14.67	280.5	127.0	12.620	8.80	4.0	2.380000	1.0
4153	1.0	106.0	1.0	0.0	0.0	196.7	100.0	33.440000	210.20	109.0	17.87	173.8	83.0	7.820	7.98	1.0	2.110000	4.0
3509	0.0	59.0	0.0	1.0	26.0	205.4	68.0	34.920000	115.00	115.0	9.78	214.7	130.0	9.660	9.40	2.0	2.540000	5.0
2820	0.0	74.0	0.0	1.0	27.0	154.1	122.0	26.200000	195.30	150.0	16.60	276.7	86.0	12.450	13.20	2.0	3.560000	4.0
1018	1.0	76.0	0.0	0.0	0.0	263.4	148.0	37.387971	230.30	69.0	19.58	170.6	101.0	7.680	11.40	5.0	3.080000	1.0
1904	1.0	90.8	0.0	1.0	33.0	167.8	91.0	28.530000	205.30	91.0	17.45	130.0	132.0	5.850	14.50	4.0	3.920000	4.0
1192	1.0	88.0	1.0	0.0	0.0	235.1	98.0	39.970000	251.80	79.0	21.40	285.9	76.0	12.870	7.20	2.0	1.940000	4.0
4980	1.0	73.0	0.0	0.0	0.0	177.2	86.6	30.120000	270.50	84.0	22.99	241.8	112.0	10.880	12.30	2.0	3.320000	3.0
1105	1.0	135.0	0.0	1.0	28.0	201.4	100.0	34.240000	246.50	117.0	20.95	154.8	131.0	6.970	12.90	4.0	3.480000	2.0
4421	1.0	67.0	0.0	0.0	0.0	177.4	102.0	30.160000	215.20	123.0	18.29	132.4	120.0	5.960	6.60	4.0	1.780000	4.0
69	1.0	150.0	0.0	0.0	0.0	178.9	101.0	30.410000	169.10	110.0	14.37	148.6	100.0	6.690	13.80	3.0	3.730000	4.0
2724	1.0	182.0	0.0	0.0	0.0	279.1	124.0	47.450000	180.50	108.0	15.34	217.5	104.0	9.790	9.50	11.0	2.570000	2.0
736	1.0	68.0	0.0	0.0	0.0	162.1	86.0	27.560000	155.00	86.0	13.18	189.7	87.0	8.540	11.00	9.0	2.970000	5.0
1240	1.0	108.0	0.0	1.0	34.0	162.1	83.0	27.560000	171.80	117.0	14.60	259.8	76.0	11.690	9.60	3.0	2.590000	4.0
3991	1.0	123.0	0.0	0.0	0.0	132.2	122.0	22.470000	169.90	101.0	14.44	150.1	123.0	6.750	12.90	11.0	3.480000	3.0
2785	1.0	38.0	0.0	0.0	0.0	175.7	109.0	29.870000	211.80	97.0	18.00	137.9	109.0	6.210	9.20	3.0	2.624881	5.0
2950	1.0	133.0	2.0	0.0	0.0	117.8	100.0	20.030000	199.20	105.0	16.93	244.1	119.0	10.980	11.80	4.0	3.190000	0.0
4137	1.0	76.0	0.0	0.0	0.0	222.0	99.0	37.740000	211.22	122.0	18.35	204.1	131.0	9.180	8.20	5.0	2.210000	3.0
2934	1.0	24.0	0.0	0.0	0.0	149.0	73.0	25.330000	131.00	81.0	11.14	238.6	69.0	10.606	8.60	3.0	2.320000	2.0
4406	1.0	116.0	1.0	0.0	0.0	232.2	98.0	39.470000	244.70	91.0	20.80	151.3	69.0	6.810	10.60	1.0	2.860000	7.0
1115	1.0	98.0	0.0	0.0	0.0	165.0	129.0	28.050000	202.60	113.0	17.22	172.3	94.0	7.750	12.50	7.0	3.380000	1.0
3369	1.0	73.0	0.0	0.0	0.0	137.6	113.0	23.390000	179.70	75.0	15.27	172.8	71.0	7.780	10.70	2.0	2.890000	1.0
1262	0.0	157.0	0.0	0.0	0.0	220.7	105.0	37.520000	119.30	127.0	10.14	165.1	113.0	7.430	11.50	7.0	3.110000	4.0
4624	1.0	178.0	0.0	0.0	0.0	225.4	106.0	38.320000	110.70	90.0	9.41	134.1	109.0	6.030	15.30	5.0	4.130000	2.0
1532	1.0	103.0	0.0	1.0	18.0	149.9	84.0	25.480000	170.90	84.0	14.53	171.5	112.0	7.720	11.50	7.0	3.110000	0.0
3147	0.0	62.0	0.0	0.0	0.0	245.3	91.0	41.700000	122.90	130.0	10.45	228.4	102.0	10.280	8.50	4.0	2.300000	4.0
2472	1.0	129.0	0.0	0.0	0.0	137.8	120.0	23.430000	225.80	110.0	19.19	145.2	95.0	6.530	10.20	6.0	2.750000	1.0
4494	0.0	137.0	0.0	0.0	0.0	248.3	63.0	42.210000	250.40	108.0	21.28	159.1	94.0	7.160	9.60	4.0	2.590000	1.0
2238	1.0	61.0	0.0	0.0	0.0	267.1	104.0	45.410000	180.40	131.0	15.33	230.6	106.0	10.380	12.10	4.0	4.670000	1.0
974	0.0	81.4	0.0	1.0	31.0	135.9	90.0	23.100000	271.00	84.0	23.04	179.1	89.0	8.060	9.50	7.0	2.570000	6.0
3601	1.0	166.0	0.0	0.0	0.0	156.8	100.0	26.660000	171.50	93.0	14.58	149.1	97.0	6.710	11.20	10.0	3.020000	1.0
91	1.0	155.0	0.0	0.0	0.0	203.4	100.0	34.580000	190.90	104.0	16.23	196.0	119.0	8.820	8.90	4.0	2.400000	0.0
2535	1.0	125.0	0.0	0.0	0.0	113.0	108.0	19.210000	169.20	107.0	14.38	156.6	61.0	7.050	9.20	5.0	2.480000	2.0
1516	1.0	68.0	0.0	0.0	0.0	213.9	112.0	36.360000	260.50	100.0	22.14	233.8	97.0	10.520	8.40	3.0	2.270000	1.0

Introduction to Data Science Assignment 2023/2024¶

1. Business understanding¶

2. Data Understanding¶

2.1 Explore the data¶

2.2 Detecting duplicates and missing values¶

2.3 Exploring Continuous / Numeric Variables¶

2.4 Exploring Nominal and Binary Variables¶

2.5 - Detecting Outliers¶

2.5.1 - Interquartile Range method¶

2.5.2 - Common outliers identification using IQR, Z-score and MAD methods¶

2.6 Bivariate Analysis¶

3 - Data Preparation¶

3.2 - Data Cleaning: missing values¶

Row removal¶

Imputation¶

KNN Imputation¶

3.3 - Data Cleaning: outliers¶

3.3 - Data Reduction: Feature Selection¶

Boruta Method¶

Recursive Feature Elimination¶

3.4 - Construct Data: Discretising¶

4. Modelling¶

4.1 Data shuffle and split¶

4.2 - Sampling¶

a) Modelling with the balanced dataset¶

a.1) Nearest neighbor¶

a.2) Bayesian classifier¶

a.3) Decision Tree¶

a.4) Tree Ensembles Methods¶

a.5) Support vector machines¶

a.6) Neural Network Classifier¶

b) Modelling with the unbalanced dataset¶

b.1) Nearest neighbor¶

b.2) Bayesian classifier¶

b.3) Decision Tree¶

b.4) Tree Ensembles Methods¶

b.6) Support vector machines¶

b.7) Neural Network Classifier¶

5- Evaluation and Main Conclusions¶

Final Recomendations¶

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
0	No	128.0	no	yes	25.0	265.1	110.0	45.07	197.4	99.0	16.78	244.7	91.0	11.01	10.0	3.0	2.70	1.0
1	No	107.0	no	yes	26.0	161.6	123.0	27.47	195.5	103.0	16.62	254.4	103.0	11.45	13.7	3.0	3.70	1.0
2	No	137.0	no	no	0.0	243.4	NaN	41.38	121.2	110.0	10.30	162.6	104.0	7.32	12.2	5.0	3.29	0.0
3	No	84.0	yes	no	0.0	299.4	71.0	50.90	61.9	88.0	5.26	196.9	89.0	8.86	6.6	7.0	1.78	2.0
4	No	75.0	yes	no	0.0	166.7	113.0	28.34	148.3	122.0	12.61	186.9	121.0	8.41	10.1	3.0	2.73	3.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4995	No	50.0	no	yes	40.0	235.7	127.0	40.07	223.0	126.0	18.96	297.5	116.0	13.39	9.9	5.0	2.67	2.0
4996	Yes	152.0	no	no	0.0	184.2	90.0	31.31	256.8	73.0	21.83	213.6	113.0	9.61	14.7	2.0	3.97	3.0
4997	No	61.0	no	no	0.0	140.6	89.0	23.90	172.8	128.0	14.69	212.4	97.0	9.56	13.6	4.0	3.67	1.0
4998	No	109.0	no	no	0.0	188.8	67.0	32.10	171.7	92.0	14.59	224.4	89.0	NaN	8.5	6.0	2.30	0.0
4999	No	86.0	no	yes	34.0	129.4	102.0	22.00	267.1	104.0	22.70	154.8	100.0	6.97	9.3	16.0	2.51	0.0

	accountlength	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
count	4951.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000	4950.000000
mean	100.238295	7.763636	180.306625	100.038788	30.629386	200.679798	100.243838	17.048293	200.465697	99.932929	9.015240	10.259010	4.432525	2.772088	1.569091
std	39.718817	13.552928	53.926625	19.844529	9.148881	50.486434	19.837380	4.300503	50.498290	19.939450	2.276812	2.763712	2.448457	0.744552	1.305279
min	1.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	73.000000	0.000000	143.700000	87.000000	24.430000	166.425000	87.000000	14.130000	167.000000	87.000000	7.510000	8.500000	3.000000	2.300000	1.000000
50%	100.000000	0.000000	180.100000	100.000000	30.600000	201.000000	101.000000	17.090000	200.550000	100.000000	9.010000	10.300000	4.000000	2.780000	1.000000
75%	127.000000	17.000000	216.200000	113.000000	36.720000	234.100000	114.000000	19.897500	234.700000	113.000000	10.560000	12.000000	6.000000	3.240000	2.000000
max	243.000000	52.000000	351.500000	165.000000	59.760000	363.700000	170.000000	30.910000	395.000000	175.000000	17.770000	20.000000	20.000000	5.400000	9.000000

	totaldayminutes	totaldaycalls	totaldaycharge
totaldayminutes	1.000000	0.003293	0.992380
totaldaycalls	0.003293	1.000000	-0.000558
totaldaycharge	0.992380	-0.000558	1.000000

	totaleveminutes	totalevecalls	totalevecharge
totaleveminutes	1.00000	0.004650	1.000000
totalevecalls	0.00465	1.000000	0.004467
totalevecharge	1.00000	0.004467	1.000000

	totalnightminutes	totalnightcalls	totalnightcharge
totalnightminutes	1.000000	0.024878	0.999999
totalnightcalls	0.024878	1.000000	0.028675
totalnightcharge	0.999999	0.028675	1.000000

	totalintlminutes	totalintlcalls	totalintlcharge
totalintlminutes	1.000000	0.018186	0.994521
totalintlcalls	0.018186	1.000000	0.022057
totalintlcharge	0.994521	0.022057	1.000000

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
count	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.000000	5000.00000	5000.000000	5000.000000	5000.00000	5000.000000	5000.00000
mean	0.141400	100.226784	0.113800	0.282400	7.760680	180.303714	100.035360	30.638822	200.634184	100.250200	17.054007	200.468532	99.92300	9.017016	10.262268	4.42888	2.771706	1.56956
std	0.348469	39.554740	0.347669	0.471905	13.499377	53.725615	19.766022	9.148733	50.291279	19.759087	4.293597	50.276816	19.85791	2.271356	2.752844	2.43862	0.741653	1.29969
min	0.000000	1.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.00000	0.000000	0.000000	0.00000	0.000000	0.00000
25%	0.000000	73.000000	0.000000	0.000000	0.000000	143.900000	87.000000	24.430000	166.800000	87.000000	14.140000	167.300000	87.00000	7.510000	8.500000	3.00000	2.309000	1.00000
50%	0.000000	100.000000	0.000000	0.000000	0.000000	180.050000	100.000000	30.600000	201.000000	101.000000	17.090000	200.500000	100.00000	9.020000	10.300000	4.00000	2.780000	1.00000
75%	0.000000	127.000000	0.000000	1.000000	16.000000	216.000000	113.000000	36.750000	233.800000	113.000000	19.900000	234.125000	113.00000	10.557000	12.000000	6.00000	3.240000	2.00000
max	1.000000	243.000000	2.000000	2.000000	52.000000	351.500000	165.000000	59.760000	363.700000	170.000000	30.910000	395.000000	175.00000	17.770000	20.000000	20.00000	5.400000	9.00000

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
245	0.0	22.0	0.0	0.0	0.0	110.3	107.0	18.75	166.5	93.0	14.15	202.3	96.0	9.10	9.5	5.0	2.57	0.0
4493	0.0	83.0	0.0	1.0	37.0	133.1	117.0	22.63	171.4	89.0	14.57	218.1	92.0	9.81	10.9	4.0	2.94	3.0
4583	0.0	104.0	0.0	0.0	0.0	246.4	108.0	41.89	205.5	102.0	17.47	208.1	125.0	9.36	14.1	8.0	3.81	2.0
2242	0.0	192.0	0.0	0.0	0.0	185.0	88.0	31.45	224.9	98.0	19.12	212.4	105.0	9.56	11.4	3.0	3.08	2.0
3407	0.0	93.0	0.0	0.0	0.0	120.6	104.0	20.50	205.5	95.0	17.47	182.5	107.0	8.21	9.6	6.0	2.59	2.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1180	0.0	72.0	0.0	0.0	0.0	207.8	92.0	35.33	195.7	110.0	16.63	184.8	124.0	8.32	13.1	4.0	3.54	0.0
3441	0.0	56.0	0.0	0.0	0.0	254.8	83.0	43.32	194.0	98.0	16.49	186.2	118.0	8.38	13.5	2.0	3.65	2.2
1344	1.0	111.0	0.0	0.0	0.0	284.4	89.0	48.35	211.5	113.0	13.35	242.8	91.0	10.93	8.4	8.0	2.27	0.0
4623	0.0	74.0	0.0	0.0	0.0	207.1	79.0	35.21	182.0	100.0	15.47	233.7	73.0	10.52	7.4	8.0	2.00	1.0
1289	0.0	129.0	0.0	0.0	0.0	98.0	99.0	16.66	240.7	62.0	20.46	254.8	123.0	11.47	10.5	4.0	2.84	0.0

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
3439	1.0	110.0	0.0	0.0	0.0	220.9	100.0	37.550	279.6	76.0	23.77	239.8	107.0	10.79	13.5	9.0	3.650	1.0
4509	0.0	128.0	0.0	1.0	40.0	272.2	112.0	46.270	239.9	111.0	20.39	257.3	124.0	11.58	11.6	2.0	3.130	1.0
945	0.0	122.0	0.0	0.0	0.0	180.0	88.0	30.600	145.0	95.4	12.33	233.7	120.0	10.52	11.5	6.0	3.110	2.0
537	0.0	190.0	0.0	1.0	22.0	166.5	93.0	28.310	183.0	92.0	15.56	121.0	102.0	5.44	8.5	3.0	2.300	0.0
4887	0.0	116.0	0.0	0.0	0.0	152.4	101.0	25.910	191.0	90.0	16.24	143.0	96.0	6.44	11.1	9.0	3.000	0.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
3619	0.0	81.0	0.0	1.0	36.0	177.6	106.0	28.020	210.2	138.0	17.87	173.6	78.0	7.81	4.3	2.0	1.160	2.0
456	0.0	60.0	0.0	0.0	0.0	98.2	88.0	16.690	180.5	69.0	15.34	223.6	69.0	10.06	9.3	2.0	2.510	2.0
3686	1.0	64.0	0.0	0.0	0.0	297.8	88.0	49.454	196.0	89.0	16.66	213.7	107.0	9.62	12.0	2.0	2.482	1.0
1677	0.0	123.0	0.0	0.0	0.0	163.1	119.0	27.730	249.4	51.0	21.20	168.2	77.0	7.57	9.0	10.0	2.430	1.0
4996	1.0	152.0	0.0	0.0	0.0	184.2	90.0	31.310	256.8	73.0	21.83	213.6	113.0	9.61	14.7	2.0	3.970	3.0

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
845	0.0	144.0	0.0	1.0	51.0	283.9	98.0	48.26	192.0	109.0	16.32	196.3	85.0	8.83	10.00	4.0	2.70	1.0
4675	0.0	52.0	0.0	0.0	0.0	165.9	122.0	28.20	212.2	107.0	18.04	202.1	94.0	9.09	8.30	3.0	2.24	1.0
1672	0.0	95.0	0.0	0.0	0.0	203.4	96.0	34.58	168.6	61.0	14.33	173.0	105.0	7.79	13.70	3.0	3.70	2.0
3424	0.0	82.0	0.0	0.0	0.0	183.6	111.0	31.21	303.5	88.0	25.80	189.8	115.0	8.54	10.30	6.0	2.78	0.0
4756	0.0	84.0	0.0	0.0	0.0	172.3	92.0	29.29	224.2	96.0	19.06	223.4	100.0	10.05	9.70	1.0	2.62	2.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2646	0.0	101.0	0.0	2.0	0.0	232.7	114.0	39.56	186.4	123.0	15.84	153.3	122.0	6.90	11.50	6.0	3.11	5.0
1637	0.0	74.0	0.0	0.0	0.0	225.2	93.0	38.28	215.1	120.0	18.28	241.8	95.0	10.88	9.74	2.0	2.46	2.0
4045	0.0	122.0	0.0	1.0	21.0	196.0	104.0	33.32	194.7	73.0	16.55	150.8	92.0	6.79	9.60	4.0	2.59	0.0
1093	0.0	210.0	0.0	0.0	0.0	104.6	121.0	17.78	149.5	71.0	12.71	255.1	67.0	11.48	6.50	8.0	1.76	2.0
570	0.0	51.0	0.0	0.0	0.0	181.5	108.0	30.86	196.9	87.0	16.74	187.2	119.0	8.42	10.30	2.0	2.78	1.0

	churn	accountlength	internationalplan	voicemailplan	numbervmailmessages	totaldayminutes	totaldaycalls	totaldaycharge	totaleveminutes	totalevecalls	totalevecharge	totalnightminutes	totalnightcalls	totalnightcharge	totalintlminutes	totalintlcalls	totalintlcharge	numbercustomerservicecalls
count	48.000000	48.000000	48.000000	48.000000	48.000000	48.000000	48.000000	48.000000	48.000000	48.000000	48.000000	48.00000	48.000000	48.000000	48.00000	48.000000	48.000000	48.000000
mean	0.812500	104.316667	0.125000	0.250000	7.604167	191.066667	101.116667	32.328083	195.739583	101.104167	16.555417	193.81875	100.562500	8.718875	10.53500	4.270833	2.875935	2.562500
std	0.394443	41.501394	0.392754	0.437595	13.614353	49.643278	17.621739	8.275923	49.482537	18.585329	4.263230	41.56264	20.267263	1.867009	2.34517	2.473172	0.683270	1.855484
min	0.000000	24.000000	0.000000	0.000000	0.000000	105.000000	63.000000	17.850000	75.300000	69.000000	6.400000	130.00000	61.000000	5.850000	5.80000	1.000000	1.570000	0.000000
25%	1.000000	71.750000	0.000000	0.000000	0.000000	154.400000	89.750000	26.252500	169.725000	85.500000	14.377500	162.55000	86.750000	7.317500	8.77500	2.000000	2.372500	1.000000
50%	1.000000	101.000000	0.000000	0.000000	0.000000	177.300000	100.000000	30.140000	191.140000	104.500000	15.790000	180.15000	101.500000	8.105000	10.20000	4.000000	2.750000	2.500000
75%	1.000000	130.000000	0.000000	0.250000	4.500000	227.100000	112.250000	37.885000	226.925000	112.250000	19.287500	228.95000	118.250000	10.305000	12.02500	5.250000	3.260000	4.000000
max	1.000000	225.000000	2.000000	1.000000	40.000000	307.100000	148.000000	52.210000	308.200000	150.000000	26.200000	285.90000	132.000000	12.870000	17.00000	11.000000	4.670000	7.000000

	Model	Data	Accuracy	Recall Score	Precision Score	F1 Score
0	KNN	Balanced	0.80	0.80	0.87	0.82
1	KNN	Unbalanced	0.89	0.42	0.77	0.54
2	NeuralNetwork	Balanced	0.80	0.86	0.80	0.82
3	NeuralNetwork	Unbalanced	0.80	0.80	0.86	0.82
4	Bayes	Balanced	0.78	0.79	0.78	0.78
5	Bayes	Unbalanced	0.87	0.87	0.87	0.87
6	SVMLinear	Balanced	0.80	0.80	0.80	0.80
7	SVMLinear	Unbalanced	0.85	0.85	0.73	0.79